目录

陶哲轩Analysis I习题的参考解答及思考(第9章)

第9章

版本

Analysis I(第3版)。

章节9.1

练习9.1.1

题目:

Let \( X \) be any subset of the real line, and let \( Y \) be a set such that \( X \subseteq Y \subseteq \overline{X} \). Show that \(\overline{Y} = \overline{X} \).

证明:

我们要证明\( \overline{Y} \subseteq \overline{X}, \overline{X} \subseteq \overline{Y} \)。

先证明\( \overline{Y} \subseteq \overline{X} \):假设\( \overline{Y} \nsubseteq \overline{X} \),即\( \exists x_0 \in \overline{Y}, x_0 \notin \overline{X} \),也就是\( x_0 \)是\( Y \)的附着点,但不是\( X \)的附着点,因为\( x_0 \)不是\( X \)的附着点,因此\( \exists \epsilon_0 > 0, \forall y \in X, |x_0 - y| > \epsilon_0 \) (1) ,因为\( x_0 \)是\( Y \)的附着点,因此\( \exists y_0 \in Y, |x_0 - y_0| \leq \dfrac{\epsilon_0}{2} \),因为\( Y \subseteq \overline{X} \),因此\( y_0 \in \overline{X} \),即\( y_0 \)是\( X \)的附着点,进而\( \exists x_1 \in X, |y_0 - x_1| \leq \dfrac{\epsilon_0}{2} \),可得\( |x_0 - x_1| = |(x_0 - y_0) + (y_0 - x_1)| \leq |x_0 - y_0| + |y_0 - x_1| \leq \epsilon_0 \),然而这和(1)矛盾,因此假设不成立,有\( \overline{Y} \subseteq \overline{X} \)。

最后证明\( \overline{X} \subseteq \overline{Y} \):假设\( \overline{X} \nsubseteq \overline{Y} \),即\( \exists x_0 \in \overline{X}, x_0 \notin \overline{Y} \),也就是\( x_0 \)是\( X \)的附着点,但不是\( Y \)的附着点,因为\( x_0 \)不是\( Y \)的附着点,因此\( \exists \epsilon_0 > 0, \forall y \in Y, |x_0 - y| > \epsilon_0 \) (2) ,因为\( x_0 \)是\( X \)的附着点,因此\( \exists y_0 \in X, |x_0 - y_0| \leq \epsilon_0 \),因为\( X \subseteq Y \),因此\( y_0 \in Y \),然而这和(2)矛盾,因此假设不成立,有\( \overline{X} \subseteq \overline{Y} \)。

综上,有\( \overline{Y} = \overline{X} \)。

证毕。

练习9.1.2

题目:

Prove Lemma 9.1.11.

Lemma 9.1.11的内容:

(Elementary properties of closures). Let \( X \) and \( Y \) be arbitrary subsets of \( \mathbf{R} \). Then \( X \subseteq \overline{X} \), \( \overline{X \cup Y} = \overline{X} \cup \overline{Y} \), and \( \overline{X \cap Y} \subseteq \overline{X} \cap \overline{Y} \). If \( X \subseteq Y \), then \( \overline{X} \subseteq \overline{Y} \).

证明:

证明\( X \subseteq \overline{X} \):

\( \forall x \in X, \forall \epsilon > 0, |x - x| \leq \epsilon \),因此\( x \)是\( X \)附着点,进而\( x \in \overline{X} \)。

综上,有\( X \subseteq \overline{X} \)。

证明\( \overline{X \cup Y} = \overline{X} \cup \overline{Y} \):

证明\( \overline{X \cup Y} \subseteq \overline{X} \cup \overline{Y} \): \( \forall x \in \overline{X \cup Y} \),有\( x \)是\( X \cup Y \)的附着点,因此\( \forall \epsilon > 0, \exists y \in X \cup Y, |x - y| \leq \epsilon \) (1) ,假设\( x \)不是\( X \)的附着点,也不是\( Y \)的附着点,即\( \exists \epsilon_1 > 0, \forall y \in X, |x - y| > \epsilon_1 \),且\( \exists \epsilon_2 > 0, \forall y \in Y, |x - y| > \epsilon_2 \),取\( \epsilon_0 = \min(\epsilon_0, \epsilon_1) \),则有\( \forall y \in X \cup Y, |x - y| > \epsilon_0 \),然而这和(1)矛盾,因此假设不成立,有\( x \)是\( X \)的附着点或\( x \)是\( Y \)的附着点,进而\( x \in \overline{X} \) 或\( x \in \overline{Y} \),也就是\( x \in \overline{X} \cup \overline{Y} \)。

证明\( \overline{X} \cup \overline{Y} \subseteq \overline{X \cup Y} \): \( \forall x \in \overline{X} \cup \overline{Y} \),有 \( x \in \overline{X} \)或\( x \in \overline{Y} \),如果\( x \in \overline{X} \),即\( x \)是\( X \)的附着点,则\( \forall \epsilon > 0, \exists y \in X, |x - y| \leq \epsilon \),由于\( y \in X \),因此也有\( y \in X \cup Y \),简而言之就是, \( \forall \epsilon > 0, \exists y \in X \cup Y, |x - y| \leq \epsilon \),也就是说\( x \)是\( X \cup Y \)的附着点,进而\( x \in \overline{X \cup Y} \), \( x \in \overline{Y} \)的情况也同理,易证\( x \in \overline{X \cup Y} \)。

综上,有\( \overline{X \cup Y} = \overline{X} \cup \overline{Y} \)。

证明\( \overline{X \cap Y} \subseteq \overline{X} \cap \overline{Y} \):

\( \forall x \in \overline{X \cap Y} \),有\( x \)是\( X \cap Y \)的附着点,因此\( \forall \epsilon > 0, \exists y \in X \cap Y, |x - y| \leq \epsilon \),因为\( y \in X \cap Y \),故有\( y \in X \)且\( y \in Y \),于是有\( \forall \epsilon > 0, \exists y \in X, |x - y| \leq \epsilon \),且\( \forall \epsilon > 0, \exists y \in Y, |x - y| \leq \epsilon \),即\( x \)是\( X \)的附着点且\( x \)是\( Y \)的附着点,因此\( x \in \overline{X} \)且\( x \in \overline{Y} \),进而\( x \in \overline{X} \cap \overline{Y} \)。

证明如果\( X \subseteq Y \),则\( \overline{X} \subseteq \overline{Y} \):

\( \forall x \in \overline{X} \),有\( x \)是\( X \)的附着点,因此\( \forall \epsilon > 0, \exists y \in X, |x - y| \leq \epsilon \),而\( X \subseteq Y \),因此有\( y \in Y \),简而言之, \( \forall \epsilon > 0, \exists y \in Y, |x - y| \leq \epsilon \),因此\( x \)是\( Y \)的附着点,进而\( x \in \overline{Y} \)。

综上,有\( \overline{X} \subseteq \overline{Y} \)。

证毕。

练习9.1.3

题目:

Prove Lemma 9.1.13. (Hint: for computing the closure of \( \mathbf{Q} \), you will need Proposition 5.4.14.)

Lemma 9.1.13的内容:

The closure of \( \mathbf{N} \) is \( \mathbf{N} \). The closure of \( \mathbf{Z} \) is \( \mathbf{Z} \). The closure of \( \mathbf{Q} \) is \( \mathbf{R} \), and the closure of \( \mathbf{R} \) is \( \mathbf{R} \). The closure of the empty set \( \emptyset \) is \( \emptyset \).

Proposition 5.4.14的内容见练习5.4.5,这里不再重复。

证明:

证明\( \overline{\mathbf{N}} = \mathbf{N} \):

根据引理9.1.11,\( \mathbf{N} \subseteq \overline{\mathbf{N}} \),我们还需要证明\( \overline{\mathbf{N}} \subseteq \mathbf{N} \): \( \forall x \in \overline{\mathbf{N}} \),假设\( x \notin \mathbf{N} \),分情况讨论:

  1. 如果\( x < 0 \),则令\( \epsilon_0 := \dfrac{x}{2} \),此时\( \forall n \in \mathbf{N}, |x - n| \geq |x| > \epsilon_0 \),即\( x \)不是\( \mathbf{N} \)的附着点,这和\( x \in \overline{\mathbf{N}} \)矛盾。
  2. 如果\( x \geq 0 \),根据练习5.4.3, \( \exists N \in \mathbf{Z}, N \leq x < N + 1 \),因为\( x \geq 0 \),因此\( N, N + 1 \geq 0 \),即\( N, N + 1 \)均是自然数,令\( \epsilon_0 := \min(|x - N|, |x - N - 1|) / 2 \),由\( x \notin \mathbf{N} \)以及\( N \leq x < N + 1 \),可得\( 0 < |x - N| < 1, 0 < |x - N - 1| < 1 \),特别的, \( \epsilon_0 \)也\( > 0 \), \( \forall n \in \mathbf{N} \),可得\( |x - n| \geq \min(|x - N|, |x - N - 1|) > \epsilon_0 \) (考虑\( n \)比\( N + 1 \)大以及比\( N \)小的情况,距离肯定更大),即\( x \)不是\( \mathbf{N} \)的附着点,这和\( x \in \overline{\mathbf{N}} \)矛盾。

综上,所有情况都矛盾,假设不成立,有\( x \in \mathbf{N} \)。

至此,我们证明了\( \mathbf{N} \subseteq \overline{\mathbf{N}}, \overline{\mathbf{N}} \subseteq \mathbf{N} \),于是有 \( \overline{\mathbf{N}} = \mathbf{N} \)。

证明\( \overline{\mathbf{Z}} = \mathbf{Z} \):

根据引理9.1.11,\( \mathbf{Z} \subseteq \overline{\mathbf{Z}} \),我们还需要证明\( \overline{\mathbf{Z}} \subseteq \mathbf{Z} \): \( \forall x \in \overline{\mathbf{Z}} \),假设\( x \notin \mathbf{Z} \):根据练习5.4.3,\( \exists N \in \mathbf{Z}, N \leq x < N + 1 \),令\( \epsilon_0 := \min(|x - N|, |x - N - 1|) / 2 \),由\( x \notin \mathbf{Z} \)以及\( N \leq x < N + 1 \),可得\( 0 < |x - N| < 1, 0 < |x - N - 1| < 1 \),特别的, \( \epsilon_0 \)也\( > 0 \), \( \forall n \in \mathbf{Z} \),可得\( |x - n| \geq \min(|x - N|, |x - N - 1|) > \epsilon_0 \),即\( x \)不是\( \mathbf{Z} \)的附着点,这和\( x \in \overline{\mathbf{Z}} \)矛盾,因此假设不成立,有\( x \in \mathbf{Z} \)。

至此,我们证明了\( \mathbf{Z} \subseteq \overline{\mathbf{Z}}, \overline{\mathbf{Z}} \subseteq \mathbf{Z} \),于是有 \( \overline{\mathbf{Z}} = \mathbf{Z} \)。

证明\( \overline{\mathbf{Q}} = \mathbf{R} \):

明显有\( \overline{\mathbf{Q}} \subseteq \mathbf{R} \)(附着点是实数),还得证明\( \mathbf{R} \subseteq \overline{\mathbf{Q}} \): \( \forall x \in \mathbf{R}, \forall \epsilon > 0 \),令\( y := x + \epsilon \),根据定理5.4.14, \( \exists q \in \mathbf{Q}, x < q < y \),进而\( |q - x| < |y - x| = \epsilon \),因此\( x \)是\( \mathbf{Q} \)的附着点,进而\( x \in \overline{\mathbf{Q}} \)。

至此,我们证明了\( \overline{\mathbf{Q}} \subseteq \mathbf{R}, \mathbf{R} \subseteq \overline{\mathbf{Q}} \),于是有 \( \overline{\mathbf{Q}} = \mathbf{R} \)。

证明\( \overline{\mathbf{R}} = \mathbf{R} \):

明显有\( \overline{\mathbf{R}} \subseteq \mathbf{R} \)(附着点是实数),根据引理9.1.11,\( \mathbf{R} \subseteq \overline{\mathbf{R}} \)。综上,\( \overline{\mathbf{R}} = \mathbf{R} \)。

证明\( \overline{\emptyset} = \emptyset \):

假设\( \overline{\emptyset} \)非空,则\( \exists x \in \overline{\emptyset} \), \( x \)是\( \emptyset \)的附着点,因此\( \forall \epsilon > 0 \),要\( \exists y \in \emptyset, |x - y| \leq \epsilon \),而 \( \emptyset \)中没有元素,不存在这样的\( y \),矛盾,因此假设不成立,\( \overline{\emptyset} = \emptyset \)。

证毕。

练习9.1.4

题目:

Give an example of two subsets \( X, Y \) of the real line such that \( \overline{X \cap Y} \neq \overline{X} \cap \overline{Y} \).

例子:

令\( X := (1, 2), Y := (2, 3) \),此时\( \overline{X \cap Y} = \emptyset \),而\( \overline{X} \cap \overline{Y} = [1, 2] \cap [2, 3] = \{ 2 \} \neq \overline{X \cap Y} \)。

练习9.1.5

题目:

Prove Lemma 9.1.14. (Hint: in order to prove one of the two implications here you will need axiom of choice, as in Lemma 8.4.5.)

Lemma 9.1.14的内容:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( x \in \mathbf{R} \). Then \( x \) is an adherent point of \( X \) if and only if there exists a sequence \( (a_n)_{n = 0}^{\infty} \), consisting entirely of elements in \( X \), which converges to \( x \).

证明:

必要性:

如果\( x \)是\( X \)附着点,\( \forall n \in \mathbf{N} \),令\( X_n := \{ y \in X : |x - y| \leq \dfrac{1}{n + 1} \} \) (取\( \dfrac{1}{n + 1} \)而非\( \dfrac{1}{n} \)是因为后者在\( n = 0 \) 未定义,另一种方法就是让序列下标从\( 1 \)开始,构造出序列后,再对下标进行位移,从而把起始下标变成\( 0 \)),令\( \epsilon := \dfrac{1}{n + 1} \),因为\( x \)是\( X \)的附着点,因此\( \exists y \in X, |x - y| \leq \epsilon = \dfrac{1}{n + 1} \),可得\( y \in X_n \),于是有\( X_n \)非空。根据选择公理,\( \exists ((a_n)_{n \in \mathbf{N}} = (a_n)_{n = 0}^{\infty}) \in \prod_{n \in \mathbf{N}} X_n \),这里\( \forall n \in \mathbf{N}, a_n \in X_n \),因此也有\( a_n \in X \),即\( (a_n)_{n = 0}^{\infty} \)完全由\( X \)中的元素组成,我们证明\( \lim_{n \to \infty} a_n = x \):因为\( \lim_{n \to \infty} \dfrac{1}{n + 1} = 0 \),因此\( \forall \epsilon > 0, \exists N \in \mathbf{N}, \forall n \geq N, |\dfrac{1}{n + 1}| = \dfrac{1}{n + 1} \leq \epsilon \),而\( a_n \in X_n \),根据\( X_n \)的定义,有\( |x - a_n| \leq \dfrac{1}{n + 1} \leq \epsilon \),综上,有\( \lim_{n \to \infty} a_n = x \)。

充分性:

如果\( \exists \)序列\( (a_n)_{n = 0}^{\infty} \) 满足\( \lim_{n \to \infty} a_n = x \)且 \( \forall n \in \mathbf{N}, a_n \in X \),因为\( \lim_{n \to \infty} a_n = x \),因此\( \forall \epsilon > 0, \exists N \in \mathbf{N}, \forall n \geq N, |a_n - x| \leq \epsilon \),特别的,有\( |a_N - x| \leq \epsilon \)且\( a_N \in X \),于是有\( x \)为\( X \)的附着点。

证毕。

练习9.1.6

题目:

Let \( X \) be a subset of \( \mathbf{R} \). Show that \( \overline{X} \) is closed (i.e., \( \overline{\overline{X}} = \overline{X} \)). Furthermore, show that if \( Y \) is any closed set that contains \( X \), then \( Y \) also contains \( \overline{X} \). Thus the closure \(\overline{X} \) of \( X \) is the smallest closed set which contains \( X \).

证明:

证明\( \overline{X} \)是封闭的,即\( \overline{\overline{X}} = \overline{X} \):

根据引理9.1.11,有\( \overline{X} \subseteq \overline{\overline{X}} \),我们还需要证明\( \overline{\overline{X}} \subseteq \overline{X} \):假设\( \overline{\overline{X}} \nsubseteq \overline{X} \),即\( \exists x \in \overline{\overline{X}}, x \notin \overline{X} \),因为\( x \notin \overline{X} \),因此\( \exists \epsilon_0 > 0, \forall y \in X, |x - y| > \epsilon_0 \) (1) ,针对\( \dfrac{\epsilon_0}{2} \),因为\( x \in \overline{\overline{X}} \),因此\( \exists y_0 \in \overline{X}, |x - y_0| \leq \dfrac{\epsilon_0}{2} \),因为\( y_0 \in \overline{X} \),因此\( \exists x_0 \in X, |y_0 - x_0| \leq \dfrac{\epsilon_0}{2} \),可得\( |x - x_0| = |x - y_0| + |y_0 - x_0| \leq \epsilon_0 \),然而\( x_0 \in X \),这违反了(1),矛盾,因此假设不成立,有\( \overline{\overline{X}} \subseteq \overline{X} \)。综上,有\( \overline{\overline{X}} = \overline{X} \)。

证明如果\( Y \)是封闭集且\( X \subseteq Y \),则\( \overline{X} \subseteq Y \):

假设\( \overline{X} \nsubseteq Y \),即 \( \exists x \in \overline{X}, x \notin Y \),因为\( Y \)是封闭集,因此\( \overline{Y} = Y \),于是也有\( x \notin \overline{Y} \),进而\( \exists \epsilon_0 > 0, \forall y \in Y, |x - y| > \epsilon_0 \) (1) ,又因为\( x \in \overline{X} \),因此\( \exists y_0 \in X, |x - y_0| \leq \epsilon_0 \),而\( X \subseteq Y \),因此\( y_0 \in Y \),这违反了(1),矛盾,因此假设不成立,有\( \overline{X} \subseteq Y \)。

证毕。

练习9.1.7

题目:

Let \( n \geq 1 \) be a positive integer, and let \( X_1, \dots, X_n \) be closed subsets of \( \mathbf{R} \). Show that \( X_1 \cup X_2 \cup \dots \cup X_n \) is also closed.

注:

题目证明的是有限个封闭集的并集仍然是封闭集,但是无限个封闭集的并集则不一定是封闭集,具体的,令\( I \)为一个集合,\( \forall \alpha \in I \),令\( X_{\alpha} \)为一个封闭集,则\( \bigcup_{\alpha \in I} X_{\alpha} \)不一定是封闭集,举个例子:令\( I := \mathbf{N} \setminus \{ 0 \} \), \( \forall n \in I \),令\( X_n := [\dfrac{1}{n}, 1] \),这里\( X_n \)均是封闭集,但\( \bigcup_{n \in I} X_n = (0, 1] \)不是封闭集(注意:这里\( 0 \)是取不到的,因为\( \dfrac{1}{n} \)永远不等于\( 0 \))。

证明:

对\( n \)进行数学归纳。

当\( n = 1 \)时,\( X_1 \cup X_2 \cup \dots \cup X_n = X_1 \)是封闭集。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时,令\( U := X_1 \cup X_2 \cup \dots \cup X_k \),我们要证明\( U \cup X_n \)是封闭集,假设\( U \cup X_n \)不是封闭集,即\( (\overline{U \cup X_n}) \neq (U \cup X_n) \) (1) ,根据引理9.1.11,有\( (U \cup X_n) \subseteq (\overline{U \cup X_n}) \),因此导致(1)的原因只可能是 \( (\overline{U \cup X_n}) \nsubseteq (U \cup X_n) \),即\( \exists x \in \overline{U \cup X_n}, x \notin U \cup X_n \),由\( x \notin U \cup X_n \),有\( x \notin U \)且\( x \notin X_n \),因为\( X_n \)是封闭集,即\( \overline{X_n} = X_n \),因此也有\( x \notin \overline{X_n} \),进而\( \exists \epsilon_0 > 0, \forall y \in X_n, |x - y| > \epsilon_0 \) (2) ,根据归纳假设,\( U \)是封闭集,即\( \overline{U} = U \),于是也有\( x \notin \overline{U} \),进而\( \exists \epsilon_1 > 0, \forall y \in U, |x - y| > \epsilon_1 \) (3) ,因为\( x \in \overline{U \cup X_n} \),因此 \( \exists y_0 \in U \cup X_n, |x - y_0| \leq \min(\epsilon_0, \epsilon_1) \),而由(2)和(3),可得\( y_0 \notin X_n \)且\( y_0 \notin U \),即\( y_0 \notin U \cup X_n \),然而这和\( y_0 \in U \cup X_n \)矛盾,因此假设不成立,有\( U \cup X_n \)是封闭集,即\( n = k + 1 \)时也成立。

至此,归纳完毕,命题成立。

证毕。

练习9.1.8

题目:

Let \( I \) be a non-empty set (possibly infinite), and for each \(\alpha \in I \) let \( X_{\alpha} \) be a closed subset of \(\mathbf{R} \). Show that the intersection \( \bigcap_{\alpha \in I} X_{\alpha} \) (defined in (3.3)) is also closed.

证明:

假设\( \bigcap_{\alpha \in I} X_{\alpha} \)不是封闭集,根据引理9.1.11,有\( \exists x \in \overline{\bigcap_{\alpha \in I} X_{\alpha}}, x \notin \bigcap_{\alpha \in I} X_{\alpha} \),因为\( x \notin \bigcap_{\alpha \in I} X_{\alpha} \),因此\( \exists \beta \in I, x \notin X_{\beta} \),因为\( X_{\beta} \)是封闭集,因此也有\( x \notin \overline{X_{\beta}} \),于是\( \exists \epsilon_0 > 0, \forall y \in X_{\beta}, |x - y| > \epsilon_0 \) (1) ,因为\( x \in \overline{\bigcap_{\alpha \in I} X_{\alpha}} \),因此\( \exists y_0 \in \bigcap_{\alpha \in I} X_{\alpha}, |x - y_0| \leq \epsilon_0 \),但是\( y_0 \)也\( \in X_{\beta} \),这和(1)矛盾,因此假设不成立,有\( \bigcap_{\alpha \in I} X_{\alpha} \)是封闭集。

证毕。

练习9.1.9

题目:

Let \( X \) be a subset of the real line. Show that every adherent point of \( X \) is either a limit point or an isolated point of \( X \), but cannot be both. Conversely, show that every limit point and every isolated point of \( X \) is an adherent point of \( X \).

证明:

正向:

\( \forall x \in \overline{X} \),根据引理9.1.14,\( \exists \)序列\( (a_n)_{n = 0}^{\infty} \) 满足\( \forall n \in \mathbf{N}, a_n \in X \)且\( \lim_{n \to \infty} a_n = x \),令\( A \)为所有满足这样条件的序列组成的集合,可得\( A \)非空:

  1. 如果\( \exists \)序列\( (b_n)_{n = 0}^{\infty} \in A \)满足 \( \forall n \in \mathbf{N}, b_n \neq x \),则根据引理9.1.14,可得 \( x \)是\( X \setminus \{ x \} \)的附着点,因此\( x \)是\( X \)的极限点。
  2. 如果\( \forall (b_n)_{n = 0}^{\infty} \in A \),均有 \( \exists n_0 \in \mathbf{N}, b_{n_0} = x \) (1) ,我们证明\( x \)是\( X \)的孤立点,假设\( x \)不是\( X \)的孤立点,由(1)可得,\( x \in X \),因此导致\( x \)不是\( X \)的孤立点的原因只可能是因为\( \forall \epsilon > 0, \exists y \in X \setminus \{ x \}, |x - y| \leq \epsilon \),于是可得\( x \)是\( X \setminus \{ x \} \)的附着点,从而根据引理9.1.14, \( \exists \)序列\( (c_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, c_n \in X \setminus \{ x \} \) 且\( \lim_{n \to \infty} c_n = x \),特别的,有\( \forall n \in \mathbf{N}, c_n \neq x \)以及\( (c_n)_{n = 0}^{\infty} \in A \),这和(1)矛盾,因此假设不成立,有\( x \)是\( X \)的孤立点。

明显情况1和情况2至少有一个成立,我们证明情况1和情况2不可能同时成立,假设情况1和情况2同时成立,则有\( x \)是\( X \)的极限点,也是\( X \)的孤立点,由\( x \)是\( X \)的孤立点,可得\( \exists \epsilon_0 > 0, \forall y \in X \setminus \{ x \}, |x - y| > \epsilon_0 \) (2) ,而由\( x \)是\( X \)的极限点,即\( x \)是\( X \setminus \{ x \} \)的附着点,可得\( \exists y_0 \in X \setminus \{ x \}, |x - y_0| \leq \epsilon_0 \),这和(2)矛盾,因此假设不成立,有情况1和情况2不能同时成立,加上情况1和情况2至少有一个成立,可得情况1和情况2有且仅有一个成立,即\( x \)是\( X \)的极限点或\( x \)是\( X \)的孤立点,但\( x \)不能同时是\( X \)的极限点以及孤立点。

反向:

\( \forall x \)为\( X \)的极限点,有\( x \)是\( X \setminus \{ x \} \)的附着点,从而\( \exists \)序列\( (a_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, a_n \in X \setminus \{ x \} \) 且\( \lim_{n \to \infty} a_n = x \),特别的,有\( \forall n \in \mathbf{N}, a_n \in X \),根据引理9.1.14,有\( x \)是\( X \)的附着点。

\( \forall x \)为\( X \)的孤立点,有\( x \in X \),而\( X \subseteq \overline{X} \),因此有\( x \in \overline{X} \),即\( x \)是\( X \)的附着点。

证毕。

练习9.1.10

题目:

If \( X \) is a non-empty subset of \(\mathbf{R} \), show that \( X \) is bounded if and only if \( \inf(X) \) and \( \sup(X) \) are finite.

证明:

必要性:

如果\( X \)有界,则\( \exists M > 0 \in \mathbf{R}, X \subseteq [-M, M] \),可得\( \forall x \in X, -M \leq x \leq M \),于是有\( M \)是\( X \)的上界以及\( -M \)是\( X \)的下界,进而有\( -M \leq \inf(X) \leq \sup(X) \leq M \),因此\( \inf(X), \sup(X) \)都是有限实数。

充分性:

如果\( \inf(X), \sup(X) \)都是有限实数,则取\( M := \max(|\inf(X)|, |\sup(X)|) \),可得\( \forall x \in X, -M \leq \inf(X) \leq x \leq \sup(X) \leq M \),从而\( X \subseteq [-M, M] \),即\( X \)有界。

证毕。

练习9.1.11

题目:

Show that if \( X \) is a bounded subset of \( \mathbf{R} \), then the closure \( \overline{X} \) is also bounded.

证明:

因为\( X \)有界,因此\( \exists M > 0, X \subseteq [-M, M] \)。

\( \forall x \in \overline{X} \):

  1. 如果\( x \in X \),则由\( X \subseteq [-M, M] \),可得\( -M \leq x \leq M \),进而可得\( -(M + 1) \leq x \leq M + 1 \),即\( x \in [-(M + 1), M + 1] \)。
  2. 如果\( x \notin X \),则由\( x \)是\( X \)的附着点,可得\( \exists y \in X, |x - y| \leq 1 \),进而可得\( y - 1 \leq x \leq y + 1 \),而由\( y \in X \),可得\( -(M + 1) \leq M \leq y \leq M \leq M + 1 \),于是有\( -(M + 1) \leq y - 1 \leq x \leq y + 1 \leq M + 1 \),即\( x \in [-(M + 1), M + 1] \)。

综上,有\( \overline{X} \subseteq [-(M + 1), M + 1] \),即\( \overline{X} \)有界。

证毕。

练习9.1.12

题目:

Show that the union of any finite collection of bounded subsets of \( \mathbf{R} \) is still a bounded set. Is this conclusion still true if one takes an infinite collection of bounded subsets of \( \mathbf{R} \)?

证明任意有限个有界集的并集仍是有界集:

这等价于证明:给定\( n \)为一个自然数,\( \forall 1 \leq i \leq n \),令\( X_i \) 为有界集,则\( \bigcup_{1 \leq i \leq n} X_i \)为有界集。

对\( n \)进行数学归纳:

当\( n = 0 \)时,\( \bigcup_{1 \leq i \leq n} X_i = \emptyset \),此时任取\( M > 0 \in \mathbf{R} \),均有\( \bigcup_{1 \leq i \leq n} X_i \subseteq [-M, M] \),即\( \bigcup_{1 \leq i \leq n} X_i \)有界。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时,根据归纳假设,\( \bigcup_{1 \leq i \leq k} X_k \)有界,因此\( \exists M_0 > 0, \forall x \in \bigcup_{1 \leq i \leq k} X_k, -M_0 \leq x \leq M_0 \),而\( X_{k + 1} \)也有界,因此\( \exists M_1 > 0, \forall x \in X_{k + 1}, -M_1 \leq x \leq M_1 \),取\( M := \max(M_0, M_1) \),有\( \forall x \in ((\bigcup_{1 \leq i \leq k} X_k) \cup X_{k + 1} = \bigcup_{1 \leq i \leq k + 1} X_k), -M \leq x \leq M \),即\( \bigcup_{1 \leq i \leq k + 1} X_k \subseteq [-M, M] \),从而\( \bigcup_{1 \leq i \leq k + 1} X_k \)有界。

至此,归纳完毕,命题成立。

证毕。

无限个有界集的并集是否一定是有界集?

该结论不成立,举个例子,令\( U := \{ \{ n \} : n \in \mathbf{Z} \} \),明显有\( \forall A \in U \),\( A \)均有界,但\( \bigcup_{A \in U} A = \mathbf{Z} \)无界。

练习9.1.13

题目:

Prove Theorem 9.1.24. (Hint: to show 1 implies 2, use the Bolzano-Weierstrass theorem (Theorem 6.6.8) and Corollary 9.1.17. To show 2 implies 1, argue by contradiction, using Corollary 9.1.17 to establish that \( X \) is closed. You will need the axiom of choice to show that \( X \) is bounded, as in Lemma 8.4.5.)

Theorem 9.1.24的内容:

(Heine-Borel theorem for the line). Let \( X \) be a subset of \( \mathbf{R} \). Then the following two statements are equivalent:

  1. \( X \) is closed and bounded.
  2. Given any sequence \( (a_n)_{n = 0}^{\infty} \) of real numbers which takes values in \( X \) (i.e., \( a_n \in X \) for all \( n \)), there exists a subsequence \( (a_{n_j})_{j = 0}^{\infty} \) of the original sequence, which converges to some number \( L \) in \( X \).

证明:

证明1 \( \Longrightarrow \) 2:

给定任意数列\( (a_n)_{n = 0}^{\infty}, \forall n \in \mathbf{N}, a_n \in X \),因为\( X \)有界,因此\( \exists M > 0, X \subseteq [-M, M] \),可得\( \forall n \in \mathbf{N}, |a_n| \leq M \),根据定理6.6.8,可得\( (a_n)_{n = 0}^{\infty} \)存在一个收敛的子序列\( (a_{n_j})_{j = 0}^{\infty} \),令\( L := \lim_{j \to \infty} a_{n_j} \),因为\( \forall j \in \mathbf{N}, a_{n_j} \in X \),由\( X \)是封闭集以及推论9.1.17,可得\( L \in X \)。

证明2 \( \Longrightarrow \) 1:

先证明\( X \)是封闭集,假设\( X \)不是封闭集,则\( \exists x_0 \in \overline{X} \setminus X \),因为\( x_0 \in \overline{X} \),因此根据引理9.1.14可得, \( \exists \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in X \)且\( \lim_{n \to \infty} a_n = x_0 \),根据2可得,\( (a_n)_{n = 0}^{\infty} \)存在收敛的子序列\( (a_{n_j})_{j = 0}^{\infty} \) 满足\( L := \lim_{j \to \infty} a_{n_j} \in X \),然而根据定理6.6.5,有\( L = x_0 \),进而得到\( x_0 \in X \),然而这和\( x_0 \notin X \)矛盾,因此假设不成立,有\( X \)是封闭集。

最后证明\( X \)是有界集,如果\( X = \emptyset \),则\( X \)明显是有界集,下面考虑\( X \neq \emptyset \)的情况,先证明\( X \)有上界,假设\( X \)没有上界,则\( \forall n \in \mathbf{N}, \exists x \in X, x > n \),令\( X_n := \{ x \in X : x > n \} \),可得\( X_n \)非空,根据选择公理,我们可以得到一个序列\( (a_n)_{n = 0}^{\infty} \) 满足\( \forall n \in \mathbf{N}, a_n \in X_n \),即\( a_n > n \),特别的,有\( (a_n)_{n = 0}^{\infty} \)为严格单调递增数列,接下来我们证明\( (a_n)_{n = 0}^{\infty} \)的任意子序列均不收敛,给定任意\( (a_n)_{n = 0}^{\infty} \)的子序列\( (a_{n_j})_{j = 0}^{\infty} \),因为\( (a_n)_{n = 0}^{\infty} \)为严格单调递增数列以及 \( \forall j \in \mathbf{N}, n_j \geq n \),可得\( a_{n_j} \geq a_n > n \),特别的,有\( (a_{n_j})_{j = 0}^{\infty} \)也无上界,进而 \( (a_{n_j})_{j = 0}^{\infty} \)也不收敛,至此,我们证明\( (a_n)_{n = 0}^{\infty} \)的任意子序列均不收敛,而\( (a_n)_{n = 0}^{\infty} \)中的元素全由\( X \)中的元素组成(因为 \( X_n \subseteq X \)),这违反了2,因此假设不成立,有\( X \)有上界,同理易证\( X \)有下界,进而由\( X \neq \emptyset \)以及定理5.5.9(和对应的下确界版本),可得\( \sup(X), \inf(X) \)存在且均是有限实数,再根据练习9.1.10,可得\( X \)有界。

证毕。

练习9.1.14

题目:

Show that any finite subset of \( \mathbf{R} \) is closed and bounded.

证明:

令\( X \)为任意\( \mathbf{R} \)的有限子集,记\( n := \#(X) \),可得\( \exists \{ 1 \leq i \leq n \} \to X \)的双射函数 \( (x_i)_{1 \leq i \leq n} \),特别的,有\( X = \bigcup_{1 \leq i \leq n} \{ x_i \} \)。

先证明\( X \)是封闭的,通过推论9.1.17,易证\( \forall 1 \leq i \leq n, \{ x_i \} \) 是封闭集,进而根据练习9.1.7,有\( X = \bigcup_{1 \leq i \leq n} \{ x_i \} \)也是封闭集。

最后证明\( X \)是有界的,\( \forall 1 \leq i \leq n, \{ x_i \} \)明显是有界集,进而根据练习9.1.12,有\( X = \bigcup_{1 \leq i \leq n} \{ x_i \} \)也是有界集。

证毕。

练习9.1.15

题目:

Let \( E \) be a bounded subset of \( \mathbf{R} \), and let \( S := \sup(E) \) be the least upper bound of \( E \). (Note from the least upper bound principle, Theorem 5.5.9, that \( S \) is a real number.) Show that \( S \) is an adherent point of \( E \), and is also an adherent point of \( \mathbf{R} \setminus E \).

证明:

先证明\( S \)是\( E \)的附着点,\( \forall \epsilon > 0 \),有\( S - \epsilon < S \),再由\( S \)是\( E \)的上确界,可得\( \exists y \in E, S - \epsilon < y \),又\( y \leq S \),可得\( S - \epsilon < y \leq S < S + \epsilon \),进而有\( |y - S| \leq \epsilon \),综上,有\( S \)是\( E \)的附着点。

最后证明\( S \)也是\( \mathbf{R} \setminus E \)的附着点, \( \forall \epsilon > 0 \),有\( S + \epsilon > S \),而\( \forall x \in E \),有\( x \leq S \),因此\( S + \epsilon \notin E \),可得\( S + \epsilon \in \mathbf{R} \setminus E \),而\( |(S + \epsilon) - S| \leq \epsilon \),综上,有\( S \)是\( \mathbf{R} \setminus E \)的附着点。

证毕。

章节9.2

练习9.2.1

题目:

Let \( f : \mathbf{R} \to \mathbf{R}, g : \mathbf{R} \to \mathbf{R}, h : \mathbf{R} \to \mathbf{R} \). Which of the following identities are true, and which ones are false? In the former case, give a proof; in the latter case, give a counterexample.

  1. \( (f + g) \circ h = (f \circ h) + (g \circ h) \)
  2. \( f \circ (g + h) = (f \circ g) + (f \circ h) \)
  3. \( (f + g) \cdot h = (f \cdot h) + (g \cdot h) \)
  4. \( f \cdot (g + h) = (f \cdot g) + (f \cdot h) \)

解答1:

该等式成立,证明如下:

首先,等式左边的函数和等式右边的函数定义域和值域均相等,除此之外,\( \forall x \in X, ((f + g) \circ h)(x) = (f + g)(h(x)) = f(h(x)) + g(h(x)) = (f \circ h)(x) + (g \circ h)(x) \),综上,有\( (f + g) \circ h = (f \circ h) + (g \circ h) \)。

证毕。

解答2:

该等式不成立,举个例子,令\( f(x) := x^2, g(x) := x^2, h(x) := x^2 \),则\( (f \circ (g + h))(1) = 4, ((f \circ g) + (f \circ h))(1) = 2 \),有\( (f \circ (g + h))(1) \neq ((f \circ g) + (f \circ h))(1) \),因此\( f \circ (g + h) \neq (f \circ g) + (f \circ h) \)。

解答3:

该等式成立,证明如下:

首先,等式左边的函数和等式右边的函数定义域和值域均相等,除此之外,\( \forall x \in X, ((f + g) \cdot h)(x) = (f + g)(x) \cdot h(x) = (f(x) + g(x)) \cdot h(x) = (f(x) \cdot h(x)) + (g(x) \cdot h(x)) = (f \cdot h)(x) + (g \cdot h)(x) \),综上,有\( (f + g) \cdot h = (f \cdot h) + (g \cdot h) \)。

证毕。

解答4:

该等式成立,证明如下:

首先,等式左边的函数和等式右边的函数定义域和值域均相等,除此之外,\( \forall x \in X, (f \cdot (g + h))(x) = f(x) \cdot (g + h)(x) = f(x) \cdot (g(x) + h(x)) = (f(x) \cdot g(x)) + (f(x) \cdot h(x)) = (f \cdot g)(x) + (f \cdot h)(x) \),综上,有\( f \cdot (g + h) = (f \cdot g) + (f \cdot h) \)。

证毕。

章节9.3

练习9.3.1

题目:

Prove Proposition 9.3.9.

Proposition 9.3.9的内容:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( L \) be a real number. Then the following two statements are logically equivalent:

  1. \( f \) converges to \( L \) at \( x_0 \) in \( E \).
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), the sequence \( (f(a_n))_{n = 0}^{\infty} \) converges to \( L \).

证明:

证明1 \( \Longrightarrow \) 2:

给定任意序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),则\( \forall \epsilon > 0 \),因为\( f \)在\( E \)中的\( x_0 \)收敛到\( L \),因此\( \exists \delta > 0, \forall x \in E, |x - x_0| < \delta, |f(x) - L| \leq \epsilon \) (1) ,因为\( \lim_{n \to \infty} a_n = x_0 \),因此\( \exists N \in \mathbf{N}, \forall n \geq N, |a_n - x_0| \leq \dfrac{\delta}{2} < \delta \),又\( a_n \in E \),从而由(1)可得\( |f(a_n) - L| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists N \in \mathbf{N}, \forall n \geq N, |f(a_n) - L| \leq \epsilon \),即序列\( (f(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \)。

证明2 \( \Longrightarrow \) 1:

\( \forall \epsilon > 0 \),假设\( \forall \delta > 0 \),均\( \exists x \in E, |x - x_0| < \delta \)使得\( |f(x) - L| > \epsilon \) (1) ,则特别的,\( \forall n \in \mathbf{N} \),令\( X_n := \{ x \in E, |x - x_0| < \dfrac{1}{n + 1}, |f(x) - L| > \epsilon \} \),由\( \dfrac{1}{n + 1} > 0 \)以及(1)可得,\( X_n \)非空,根据选择公理,可以得到序列\( (a_n)_{n = 0}^{\infty}, \forall n \in \mathbf{N}, a_n \in X_n \),也就是\( a_n \in E, |a_n - x_0| < \dfrac{1}{n + 1}, |f(a_n) - L| > \epsilon \) (2) ,易证\( \lim_{n \to \infty} a_n = x_0 \),则根据2(注意,是序号2不是序号(2)),有序列\( f(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \),然而由(2)可知,\( \forall n \in \mathbf{N}, |f(a_n) - L| > \epsilon \),可得\( f(a_n))_{n = 0}^{\infty} \)不收敛于\( L \),矛盾,因此假设不成立,有\( \exists \delta > 0, \forall x \in E, |x - x_0| < \delta, |f(x) - L| \leq \epsilon \)。综上,\( f \)在\( E \)中的\( x_0 \)收敛到\( L \)。

证毕。

练习9.3.2

题目:

Prove the remaining claims in Proposition 9.3.14.

Proposition 9.3.14的内容:

(Limit laws for functions). Let \( X \) be a subset of \( \mathbf{R} \), let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( f: X \to \mathbf{R} \) and \( g: X \to \mathbf{R} \) be functions. Suppose that \( f \) has a limit \( L \) at \( x_0 \) in \( E \), and \( g \) has a limit \( M \) at \( x_0 \) in \( E \). Then \( f + g \) has a limit \( L + M \) at \( x_0 \) in \( E \), \( f − g \) has a limit \( L − M \) at \( x_0 \) in \( E \), \( \max(f, g) \) has a limit \( \max(L, M) \) at \( x_0 \) in \( E \), \( \min(f, g) \) has a limit \( \min(L, M) \) at \( x_0 \) in \( E \) and \( fg \) has a limit \( LM \) at \( x_0 \) in \( E \). If \( c \) is a real number, then \( cf \) has a limit \( cL \) at \( x_0 \) in \( E \). Finally, if \( g \) is non-zero on \( E \) (i.e., \( g(x) \neq 0 \) for all \( x \in E \)) and \( M \) is non-zero, then \( f/g \) has a limit \( L/M \) at \( x_0 \) in \( E \).

注:

书中已经证明了\( f + g \)的情况,这里仅证明剩余的。

证明:

\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),因为\( f \)在\( E \)中的\( x_0 \)有极限\( L \),根据定理9.3.9,有\( (f(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \),同理,有\( (g(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( M \):

根据定理6.1.19的命题4,有\( (f(a_n) - g(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L - M \),再次根据定理9.3.9,有\( f - g \)在\( E \)中的\( x_0 \)有极限\( L - M \)。

根据定理6.1.19的命题7,有\( (\max(f(a_n), g(a_n)))_{n = 0}^{\infty} \)收敛且收敛到\( \max(L, M) \),再次根据定理9.3.9,有\( \max(f, g) \)在\( E \)中的\( x_0 \)有极限\( \max(L, M) \)。

根据定理6.1.19的命题8,有\( (\min(f(a_n), g(a_n)))_{n = 0}^{\infty} \)收敛且收敛到\( \min(L, M) \),再次根据定理9.3.9,有\( \min(f, g) \)在\( E \)中的\( x_0 \)有极限\( \min(L, M) \)。

根据定理6.1.19的命题2,有\( (f(a_n) g(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( LM \),再次根据定理9.3.9,有\( fg \)在\( E \)中的\( x_0 \)有极限\( LM \)。

令\( c \)为任意实数,根据定理6.1.19的命题3,有\( (cf(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( cL \),再次根据定理9.3.9,有\( cf \)在\( E \)中的\( x_0 \)有极限\( cL \)。

如果\( \forall x \in E, g(x) \neq 0 \)且\( M \neq 0 \),则\( \forall n \in \mathbf{N}, g(a_n) \neq 0 \),此时根据定理6.1.19的命题6,有\( (f(a_n) / g(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L/M \),再次根据定理9.3.9,有\( f/g \)在\( E \)中的\( x_0 \)有极限\( L/M \)。

证毕。

练习9.3.3

题目:

Prove Lemma 9.3.18.

Lemma 9.3.18的内容:

(Limits are local). Let \( X \) be a subset of \( \mathbf{R} \), let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), let \( f: X \to \mathbf{R} \) be a function, and let \( L \) be a real number. Let \( \delta > 0 \). Then we have \( \lim_{x \to x_0; x \in E} f(x) = L \) if and only if \( \lim_{x \to x_0; x \in E \cap (x_0 - \delta, x_0 + \delta)} f(x) = L \).

证明:

必要性:

如果\( \lim_{x \to x_0; x \in E} f(x) = L \),则\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),根据定理9.3.9,有\( (f(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \) (1) ,而\( \forall \)序列\( (b_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, b_n \in E \cap (x_0 - \delta, x_0 + \delta) \)且\( \lim_{n \to \infty} b_n = x_0 \),有\( \forall n \in \mathbf{N}, b_n \in E \)且\( \lim_{n \to \infty} b_n = x_0 \),于是根据(1),有\( (f(b_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \),再次根据定理9.3.9,有\( \lim_{x \to x_0; x \in E \cap (x_0 - \delta, x_0 + \delta)} f(x) = L \)。

充分性:

如果\( \lim_{x \to x_0; x \in E \cap (x_0 - \delta, x_0 + \delta)} f(x) = L \),则\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \cap (x_0 - \delta, x_0 + \delta) \) 且\( \lim_{n \to \infty} a_n = x_0 \),根据定理9.3.9,有\( (f(a_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \) (2) ,而\( \forall \)序列\( (b_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, b_n \in E \)且\( \lim_{n \to \infty} b_n = x_0 \),由\( \lim_{n \to \infty} b_n = x_0 \),可得\( \exists N \in \mathbf{N}, \forall n \geq N, |b_n - x_0| \leq \dfrac{\delta}{2} \),特别的,有\( \forall n \geq N, b_n \in (x_0 - \delta, x_0 + \delta) \) (3) ,构造序列\( (b'_n)_{n = 0}^{\infty}, \forall n \in \mathbf{N} \),如果\( 0 \leq n < N \),则令\( b'_n := b_N \) (4) ,否则,令\( b'_n := b_n \),可得\( (b'_n)_{n = 0}^{\infty} \)也收敛到\( x_0 \) (因为该序列和\( (b_n)_{n = 0}^{\infty} \)只有前\( N \)项可能不同),而根据(3)和(4)可得,\( \forall n \in \mathbf{N}, b'_n \in E \cap (x_0 - \delta, x_0 + \delta) \),进而由(2)以及\( \lim_{n \to \infty} b'_n = x_0 \),可得\( (f(b'_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \),进而可得\( (f(b_n))_{n = 0}^{\infty} \)收敛且收敛到\( L \)(因为这两个序列只有前\( N \)项可能不同),再次根据定理9.3.9,有\( \lim_{x \to x_0; x \in E} f(x) = L \)。

证毕。

练习9.3.4

题目:

Propose a definition for limit superior \( \lim \sup_{x \to x_0; x \in E} f(x) \) and limit inferior \( \lim \inf_{x \to x_0; x \in E} f(x) \), and then propose an analogue of Proposition 9.3.9 for your definition. (For an additional challenge: prove that analogue.)

思路:

参考了 Functions from topological spaces to complete latticesLimit Superior and Limit Inferior of Functions 以及What is the limit infimum and limit supremum of a function?中George Mathew的回答。

回想下序列的上极限,我们是不断去掉序列头部的第一个元素,从而不断获得一个更小的序列上界,直到序列上界不能再小为止(即一堆\( \sup \)组成的集合的\( \inf \)),那函数的上极限也同理,我们希望去掉一些元素,从而不断获得一个更小的函数上界,但函数不像序列,没有头部第一个元素的概念,所以我们要去掉哪些元素?答案是我们可以缩小区间,具体的,给定一个\( \delta > 0 \),一开始我们考虑区间\( (x_0 - \delta, x_0 + \delta) \)(严格讲应该是\( E \cap (x_0 - \delta, x_0 + \delta) \))对应的函数值组成的集合的上极限,然后我们缩小区间,取另外一个满足\( \delta' < \delta \)的正实数,考虑区间\( (x_0 - \delta', x_0 + \delta') \)对应的函数值组成的集合的上极限,就这样一步步缩小区间(即\( \delta \to 0 \)),最终得到上极限的极限值,下极限同理。

解答:

\( \forall \delta > 0 \),令\( F_{\delta} := \{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \} \),因为\( x_0 \)是\( E \)的附着点, 因此\( F_{\delta} \)非空,除此之外,根据定义6.2.6以及定义5.5.10,\( \sup(F_{\delta}), \inf(F_{\delta}) \) 均是良好定义的且是拓展实数(可能是\( +\infty, -\infty \)),构造函数\( s: (0, +\infty) \to \mathbf{R}, \forall \delta > 0 \),令\( s(\delta) := \sup(F_{\delta}) \),构造函数\( i: (0, +\infty) \to \mathbf{R}, \forall \delta > 0 \),令\( i(\delta) := \inf(F_{\delta}) \),根据前面的讨论,可知函数\( s, i \)均是良好定义的。

上极限\( \lim \sup_{x \to x_0; x \in E} f(x) \)的定义:

先给一个不可行的定义:

定义\( \lim \sup_{x \to x_0; x \in E} f(x) := \lim_{\delta \to 0; \delta \in (0, \mathbf{R})} s(\delta) \),可得\( \lim \sup_{x \to x_0; x \in E} f(x) = \lim_{\delta \to 0; \delta \in (0, \mathbf{R})} \sup(F_{\delta}) = \lim_{\delta \to 0; \delta \in (0, \mathbf{R})} \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)。

该定义为什么不可行?因为\( s(\delta) = \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)是拓展实数,不一定是实数,然而函数极限的定义中(定义9.3.6),很多操作不适用于拓展实数中的非实数(比如相减取绝对值),因此如果涉及非实数,则\( \lim_{\delta \to 0; \delta \in (0, \mathbf{R})} s(\delta) \)是未定义的,不过考虑到随着\( \delta \)变小,集合\( \{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \} \) 中的元素也会减少,进而\( s(\delta) = \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)也会变小,因此\( \lim_{\delta \to 0; \delta \in (0, \mathbf{R})} \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \) 实际上和\( \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) \)相等(考虑下\( \delta \)比较大的情况,这时候\( \{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \} \)元素比较多,因此\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)也会比较大,故仅有\( \delta \)较小的情况(即\( \delta \to 0 \))会影响\( \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) \)的值)。

下面给一个可行的定义:

定义\( \lim \sup_{x \to x_0; x \in E} f(x) := \inf(\{ s(\delta) : \delta > 0 \}) \),可得\( \lim \sup_{x \to x_0; x \in E} f(x) = \inf(\{ \sup(F_{\delta}) : \delta > 0 \}) = \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) \)。

下极限\( \lim \inf_{x \to x_0; x \in E} f(x) \)的定义:

定义\( \lim \inf_{x \to x_0; x \in E} f(x) := \sup(\{ i(\delta) : \delta > 0 \}) \),可得\( \lim \inf_{x \to x_0; x \in E} f(x) = \sup(\{ \inf(F_{\delta}) : \delta > 0 \}) = \sup(\{ \inf(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) \)。

Proposition 9.3.9的上极限版本:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( L^+ \) be an extended real number. Then the following two statements are logically equivalent:

  1. The limit superior of \( f \) at \( x_0 \) in \( E \) is \( L^+ \).
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), we have \( \lim \sup_{n \to \infty} f(a_n) \leq L^+ \). Furthermore, there exists a sequence \( (b_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), and \( \lim_{n \to \infty} f(b_n) = L^+ \) if \( L^+ \) is a real number, \( \lim \sup_{n \to \infty} f(b_n) = L^+ \) if \( L^+ \) is not a real number (注意,在\( L^+ \)是实数的情况下,有\( \lim_{n \to \infty} f(b_n) = L^+ \),此结论蕴含\( \lim \sup_{n \to \infty} f(b_n) = L^+ \),因为序列收敛时,上下极限均和极限相等。可以将结论统一成\( \lim \sup_{n \to \infty} f(b_n) = L^+ \),此时1和2仍然等价,这样虽然结论更弱,但形式更漂亮,具体看2 \( \Longrightarrow \) 1的证明中的注释).

Proposition 9.3.9的上极限版本的证明:

注1:需要注意下\( L^+ \)可能不是实数,在不是实数的情况下不能使用算术操作。

注2:注意下证明中序号1和序号(1)是不同的,以此类推。

证明1 \( \Longrightarrow \) 2:

如果1成立,即\( \lim \sup_{x \to x_0; x \in E} f(x) = L^+ \),即\( \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) = L^+ \):

\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),我们要证明\( \lim \sup_{n \to \infty} f(a_n) \leq L^+ \),这等价于,\( \forall N \in \mathbf{N} \),令\( (f(a_N))^+ := \sup(\{ f(a_n) : n \geq N \}) \),我们去证明\( \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \leq L^+ \), \( \forall \delta > 0 \),因为\( \lim_{n \to \infty} a_n = x_0 \),因此\( \exists N_0 \in \mathbf{N}, |a_n - x_0| \leq \dfrac{\delta}{2} < \delta \),特别的,有\( a_n \in (x_0 - \delta, x_0 + \delta) \),又\( \forall n \geq N_0, a_n \in E \),因此有\( \forall n \geq N_0, a_n \in E \cap (x_0 - \delta, x_0 + \delta) \),综合可得\( \{ a_n : n \geq N_0 \} \subseteq \{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \} \),进而可得\( (f(a_{N_0}))^+ = \sup(\{ f(a_n) : n \geq N_0 \}) \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \) (子集的上确界\( \leq \)父集的上确界),又\( \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \leq (f(a_{N_0}))^+ \),因此有\( \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)。综上,有\( \forall \delta > 0, \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \),即\( \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \)是\( \{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \} \)的下界,于是有\( \inf(\{ f(a_N)^+ : N \in \mathbf{N} \}) \leq L^+ \)。至此,我们证明了\( \lim \sup_{n \to \infty} f(a_n) \leq L^+ \)。

还得证明满足条件的序列\( (b_n)_{n = 0}^{\infty} \)的存在:

如果\( L^+ = +\infty \):此时\( \forall \delta > 0 \),有\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \geq L^+ = +\infty \),进而\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) = +\infty \)。 \( \forall n \in \mathbf{N} \),令\( \epsilon_n := \dfrac{1}{n + 1} \),因为\( \sup(\{ f(x) : x \in E \cap (x_0 - \epsilon_n, x_0 + \epsilon_n) \}) = +\infty \)(即集合无上界),因此\( \exists x_{\epsilon_n} \in E \cap (x_0 - \epsilon_n, x_0 + \epsilon_n), f(x_{\epsilon_n}) > n \)。于是令\( X_n := \{ x \in E \cap (x_0 - \epsilon_n, x_0 + \epsilon_n) : f(x) > n \} \),有\( X_n \)非空,使用选择公理,可获得序列\( (b_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, b_n \in X_n \)。先证明\( \lim_{n \to \infty} b_n = x_0 \), \( \forall \epsilon > 0, \exists N \in \mathbf{N} \setminus \{ 0 \} \)满足\( 0 < \dfrac{1}{N} < \epsilon \),于是\( \forall n \geq N \),有\( 0 < \dfrac{1}{n} \leq \dfrac{1}{N} < \epsilon \),又\( \epsilon_n = \dfrac{1}{n + 1} < \dfrac{1}{n} \),因此\( \epsilon_n < \epsilon \),由\( b_n \in X_n \),可得\( b_n \in E \cap (x_0 - \epsilon_n, x_0 + \epsilon_n) \),因此\( |b_n - x_0| < \epsilon_n < \epsilon \),综上,有\( \lim_{n \to \infty} b_n = x_0 \)。接着证明\( \lim \sup_{n \to \infty} f(b_n) = L^+ = +\infty \),易得\( \forall N \in \mathbf{N}, (f(b_n))_{n = N}^{\infty} \)均无上界(因为\( f(b_n) > n \)),即\( \sup(f(b_n))_{n = N}^{\infty} = +\infty \),于是可得\( \lim \sup_{n \to \infty} f(b_n) = L^+ = +\infty = L^+ \)。

如果\( L^+ = -\infty \):因为\( x_0 \)是\( E \)的附着点,因此 \( \exists \)序列\( (b_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, b_n \in E \) 且\( \lim_{n \to \infty} b_n = x_0 \),而根据前面的证明,有\( \lim \sup_{n \to \infty} f(b_n) \leq L^+ = -\infty \),于是可得\( \lim \sup_{n \to \infty} f(b_n) = -\infty = L^+ \)。

下面考虑\( L^+ \)是实数的情况:

先证明\( \forall \epsilon, \exists \delta > 0 \)满足 \( \forall x \in E \cap (x_0 - \delta, x_0 + \delta), f(x) < L^+ + \epsilon \) (1) : \( \forall \epsilon > 0 \),因为\( L^+ + \epsilon > L^+ \),因此\( \exists \delta > 0 \)满足\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) < L^+ + \epsilon \),进而可得\( \forall x \in E \cap (x_0 - \delta, x_0 + \delta), f(x) \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) < L^+ + \epsilon \)。

接着证明\( \forall \epsilon, \delta > 0, \exists x_{\delta} \in E \cap (x_0 - \delta, x_0 + \delta) \) 满足\( L^+ - \epsilon < f(x_{\delta}) \) (2) : \( \forall \epsilon, \delta > 0 \),因为\( L^+ \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \)以及 \( L^+ - \epsilon < L^+ \),因此\( L^+ - \epsilon < \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \),可得\( \exists x_{\delta} \in E \cap (x_0 - \delta, x_0 + \delta), L^+ - \epsilon < f(x_{\delta}) \)。

回去证明序列\( (b_n)_{n = 0}^{\infty} \)的存在, \( \forall n \in \mathbf{N} \),令\( \epsilon_n := \dfrac{1}{n + 1} \),根据(1),有\( \exists \delta_{n'} > 0 \)满足 \( \forall x \in E \cap (x_0 - \delta_{n'}, x_0 + \delta_{n'}), f(x) < L^+ + \epsilon_n \),令\( \delta_n := \min(\delta_{n'}, \dfrac{1}{n + 1}) \),有\( \delta_n \leq \delta_{n'} \),因此\( E \cap (x_0 - \delta_n, x_0 + \delta_n) \subseteq E \cap (x_0 - \delta_{n'}, x_0 + \delta_{n'}) \),进而有\( \forall x \in E \cap (x_0 - \delta_n, x_0 + \delta_n), f(x) < L^+ + \epsilon_n \) (3) ,接着根据(2),有\( \exists x_{\delta_n} \in E \cap (x_0 - \delta_n, x_0 + \delta_n) \)满足 \( L^+ - \epsilon_n < f(x_{\delta_n}) \),再由(3),有\( f(x_{\delta_n}) < L^+ + \epsilon_n \),至此,有\( |f(x_{\delta_n}) - L^+| < \epsilon_n \),综上,令\( X_n := \{ x \in E \cap (x_0 - \delta_n, x_0 + \delta_n) : |f(x) - L^+| < \epsilon_n \} \),有\( X_n \)非空(前面的\( x_{\delta_n} \in X_n \)),使用选择公理,可获得序列\( (b_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, b_n \in X_n \)。我们先证明\( \lim_{n \to \infty} b_n = x_0 \), \( \forall \epsilon > 0, \exists N \in \mathbf{N} \setminus \{ 0 \} \)使得\( 0 < \dfrac{1}{N} < \epsilon \),于是\( \forall n \geq N \),有\( 0 < \dfrac{1}{n} \leq \dfrac{1}{N} < \epsilon \),而\( \delta_n \leq \dfrac{1}{n + 1} < \dfrac{1}{n} \),因此有\( \delta_n < \epsilon \),进而可得\( (x_0 - \delta_n, x_0 + \delta_n) \subseteq (x_0 - \epsilon, x_0 + \epsilon) \),由\( b_n \in X_n \),可得\( b_n \in E \cap (x_0 - \delta_n, x_0 + \delta_n) \),进而可得\( b_n \in (x_0 - \epsilon, x_0 + \epsilon) \),即 \( |b_n - x_0| < \epsilon \),综上,有\( \lim_{n \to \infty} b_n = x_0 \)。最后,证明\( \lim_{n \to \infty} f(b_n) = L^+ \), \( \forall \epsilon > 0, \exists N \in \mathbf{N} \setminus \{ 0 \} \)使得\( 0 < \dfrac{1}{N} < \epsilon \),针对\( \forall n \geq N \),我们知道\( \epsilon_n = \dfrac{1}{n + 1} < \dfrac{1}{n} \),又\( \dfrac{1}{n} \leq \dfrac{1}{N} < \epsilon \),可得\( \epsilon_n < \epsilon \),由\( b_n \in X_n \),可得\( |f(b_n) - L^+| < \epsilon_n \),进而有\( |f(b_n) - L^+| < \epsilon \),综上,有\( \lim_{n \to \infty} f(b_n) = L^+ \)。

证明2 \( \Longrightarrow \) 1:

如果2成立,令\( L := \lim \sup_{x \to x_0; x \in E} f(x) = \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) \),我们要证明\( L = L^+ \):

根据2,\( \exists \)序列\( (b_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, b_n \in E, \lim_{n \to \infty} b_n = x_0 \) 且\( \lim \sup_{n \to \infty} f(b_n) = L^+ \) ( 注: 虽然在\( L^+ \)为实数时,我们可以得到一个比 \( \lim \sup_{n \to \infty} f(b_n) = L^+ \)更强的结论\( \lim_{n \to \infty} f(b_n) = L^+ \),但我们下面的证明不需要用到这么强的结论,这也是为什么2中\( L^+ \)为实数或者非实数的结论可以统一成 \( \lim \sup_{n \to \infty} f(b_n) = L^+ \)的原因,如果选择该统一、形式更漂亮的结论,那1 \( \Longrightarrow \) 2对应修改成证明\( \lim \sup_{n \to \infty} f(b_n) = L^+ \)就行,而不要证明更强的\( \lim_{n \to \infty} f(b_n) = L^+ \))。

如果\( L^+ = +\infty \):因为\( \lim_{n \to \infty} b_n = x_0 \),因此\( \forall \delta > 0, \exists N_{\delta} \in \mathbf{N}, \forall n \geq N_{\delta}, |b_n - x_0| \leq \dfrac{\delta}{2} < \delta \),此时由\( \lim \sup_{n \to \infty} f(b_n) = L^+ = +\infty \),可得\( (f(b_n))_{n = 0}^{\infty} \)无上界,因此\( \forall K \in \mathbf{N}, \exists n_K \geq N_{\delta}, f(b_{n_K}) \geq K \),由\( n_K \geq N_{\delta} \),可得\( |b_{n_K} - x_0| < \delta \),又\( b_{n_K} \in E \),可得\( b_{n_K} \in E \cap (x_0 - \delta, x_0 + \delta) \),进而可得\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \geq f(b_{n_K}) \),注意到,\( K \)可以取任意大,此时\( f(b_{n_K}) \geq K \)也会随之变得任意大,于是可得\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) = +\infty \)。综上,\( \forall \delta > 0 \),均有\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) = +\infty \),因此\( L = \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) = +\infty = L^+ \)。

如果\( L^+ = -\infty \):我们证明\( \forall K \in \mathbf{N}, \exists \delta_K > 0 \)满足 \( \sup(\{ f(x) : x \in E \cap (x_0 - \delta_K, x_0 + \delta_K) \}) \leq -K \),假设\( \exists K_0 \in \mathbf{N}, \forall \delta > 0 \), \( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) > -K_0 \) (1) ,此时\( \forall n \in \mathbf{N} \),令\( \delta_n := \dfrac{1}{n + 1} \),由(1),可得\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta_n, x_0 + \delta_n) \}) > -K_0 \),因此\( \exists x_{\delta_n} \in E \cap (x_0 - \delta_n, x_0 + \delta_n), f(x_{\delta_n}) > -K_0 \),于是令\( X_n := \{ x \in E \cap (x_0 - \delta_n, x_0 + \delta_n) : f(x) > -K_0 \} \),有\( X_n \)非空,使用选择公理,可以得到序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in X_n \),特别的,有\( a_n \in E \),类似于前面,易证\( \lim_{n \to \infty} a_n = x_0 \),再次根据\( \forall n \in \mathbf{N}, a_n \in X_n \),可得\( f(a_n) > -K_0 \),因此\( \lim \sup_{n \to \infty} f(a_n) \geq -K_0 > -\infty = L^+ \),然而这和2矛盾,因此假设不成立,有\( \forall K \in \mathbf{N}, \exists \delta_K > 0 \)满足 \( \sup(\{ f(x) : x \in E \cap (x_0 - \delta_K, x_0 + \delta_K) \}) \leq -K \),此时易证\( \{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \} \)无下界,进而可得\( L = \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) = -\infty = L^+ \)。

下面考虑\( L^+ \)是实数的情况:

先证明\( \forall \epsilon, \exists \delta > 0 \)满足 \( \forall x \in E \cap (x_0 - \delta, x_0 + \delta), f(x) < L^+ + \epsilon \) (2) :假设\( \exists \epsilon_0, \forall \delta > 0 \),均有\( \exists x_1 \in E \cap (x_0 - \delta, x_0 + \delta), f(x_1) \geq L^+ + \epsilon_0 \) (3) ,则我们可以构造一个序列违反2,具体的,\( \forall n \in \mathbf{N} \),令\( \delta_n := \dfrac{1}{n + 1} \),根据(3),\( \exists x_{\delta_n} \in E \cap (x_0 - \delta_n, x_0 + \delta_n), f(x_1) \geq L^+ + \epsilon_0 \),令\( X_n := \{ x \in E \cap (x_0 - \delta_n, x_0 + \delta_n) : f(x) \geq L^+ + \epsilon_0 \} \)非空,使用选择公理,可以得到序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in X_n \),特别的,有\( a_n \in E \),类似于前面,易证\( \lim_{n \to \infty} a_n = x_0 \),再次根据\( \forall n \in \mathbf{N}, a_n \in X_n \),可得\( f(a_n) \geq L^+ + \epsilon_0 \),因此\( \lim \sup_{n \to \infty} f(a_n) \geq L^+ + \epsilon_0 > L^+ \),然而这和2矛盾。

接着证明\( \forall \epsilon, \delta > 0, \exists x_{\delta} \in E \cap (x_0 - \delta, x_0 + \delta) \) 满足\( L^+ - \epsilon < f(x_{\delta}) \) (4) : \( \forall \epsilon, \delta > 0 \),因为\( \lim_{n \to \infty} b_n = x_0 \),因此\( \exists N_{\delta} \in \mathbf{N} \)满足 \( \forall n \geq N_{\delta}, |b_n - x_0| \leq \dfrac{\delta}{2} < \delta \) (5) ,由\( \lim \sup_{n \to \infty} f(b_n) = L^+ \)以及\( L^+ - \epsilon < L^+ \),可得\( \exists n_{\delta} \geq N_{\delta}, L^+ - \epsilon < f(b_{n_{\delta}}) \),由(5)可得,\( |b_{n_{\delta}} - x_0| < \delta \),又\( b_{n_{\delta}} \in E \),可得\( b_{n_{\delta}} \in E \cap (x_0 - \delta, x_0 + \delta) \),令\( x_{\delta} := b_{n_{\delta}} \),有\( x_{\delta} \)满足条件。

回去证明\( L = L^+ \)的存在:

\( \forall \epsilon > 0 \),根据(2),\( \exists \delta > 0 \)满足 \( \forall x \in E \cap (x_0 - \delta, x_0 + \delta), f(x) < L^+ + \epsilon \),于是有\( \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \leq L^+ + \epsilon \),于是可得\( L = \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \} \leq L^+ + \epsilon \),由于\( \epsilon \)可以任意小,因此可得\( L \leq L^+ \)。

\( \forall \epsilon > 0, \forall \delta > 0 \),根据(4), \( \exists x_{\delta} \in E \cap (x_0 - \delta, x_0 + \delta) \) 满足\( L^+ - \epsilon < f(x_{\delta}) \),而\( f(x_{\delta}) \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \),可得\( L^+ - \epsilon \leq \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) \),由于该不等式对任意\( \delta \)都成立,因此有\( L^+ - \epsilon \leq \inf(\{ \sup(\{ f(x) : x \in E \cap (x_0 - \delta, x_0 + \delta) \}) : \delta > 0 \}) = L \),又因为\( \epsilon \)可以任意小,因此可得\( L^+ \leq L \)。

综上,\( L \leq L^+, L^+ \leq L \),可得\( L = L^+ \)。

证毕。

Proposition 9.3.9的下极限版本:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( L^- \) be an extended real number. Then the following two statements are logically equivalent:

  1. The limit inferior of \( f \) at \( x_0 \) in \( E \) is \( L^- \).
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), we have \( \lim \inf_{n \to \infty} f(a_n) \geq L^- \). Furthermore, there exists a sequence \( (b_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), and \( \lim_{n \to \infty} f(b_n) = L^- \) if \( L^- \) is a real number, \( \lim \inf_{n \to \infty} f(b_n) = L^- \) if \( L^- \) is not a real number.

Proposition 9.3.9的下极限版本的证明:

和上极限版本类似,由于对称性,基本上就是把\( \sup, \inf \)对应改成 \( \inf, \sup \)等等,这里省略。

练习9.3.5

题目:

(Continuous version of squeeze test) Let \( X \) be a subset of \( \mathbf{R} \), let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( f: X \to \mathbf{R}, g: X \to \mathbf{R}, h: X \to \mathbf{R} \) be functions such that \( f(x) \leq g(x) \leq h(x) \) for all \( x \in E \). If we have \( \lim_{x \to x_0; x \in E} f(x) = \lim_{x \to x_0; x \in E} h(x) = L \) for some real number \( L \), show that \( \lim_{x \to x_0; x \in E} g(x) = L \).

证明:

\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),因为\( \lim_{x \to x_0; x \in E} f(x) = \lim_{x \to x_0; x \in E} h(x) = L \),根据定理9.3.9,有\( \lim_{n \to \infty} f(a_n) = L, \lim_{n \to \infty} h(a_n) = L \),又\( \forall n \in \mathbf{N}, f(a_n) \leq g(a_n) \leq h(a_n) \),根据推论6.4.14,有\( \lim_{n \to \infty} g(a_n) = L \),再次根据定理9.3.9,有\( \lim_{x \to x_0; x \in E} g(x) = L \)。

证毕。

章节9.4

练习9.4.1

题目:

Prove Proposition 9.4.7. (Hint: this can largely be done by applying the previous propositions and lemmas. Note that to prove 1, 2, and 3 are equivalent, you do not have to prove all six implications, but you do have to prove at least three; for instance, showing that 1 implies 2, 2 implies 3, and 3 implies 1 will suffice, although this is not necessarily the shortest or simplest way to do this question.)

Proposition 9.4.7的内容:

Proposition 9.4.7 (Equivalent formulations of continuity). Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, and let \( x_0 \) be an element of \( X \). Then the following four statements are logically equivalent:

  1. \( f \) is continuous at \( x_0 \).
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) consisting of elements of \( X \) with \( \lim_{n \to \infty} a_n = x_0 \), we have \( \lim_{n \to \infty} f(a_n) = f(x_0) \).
  3. For every \( \epsilon > 0 \), there exists a \( \delta > 0 \) such that \( |f(x) − f(x_0)| < \epsilon \) for all \( x \in X \) with \( |x − x_0| < \delta \).
  4. For every \( \epsilon > 0 \), there exists a \( \delta > 0 \) such that \( |f(x) − f(x_0)| \leq \epsilon \) for all \( x \in X \) with \( |x − x_0| \leq \delta \).

证明:

由\( X \subseteq \overline{X} \)以及\( x_0 \in X \),可得\( x_0 \)是\( X \)的附着点。

证明\( 1 \Longrightarrow 2 \):

因为\( f \)在\( x_0 \)连续,因此\( \lim_{x \to x_0; x \in X} f(x) = f(x_0) \),进而根据定理9.3.9,有2成立。

证明\( 2 \Longrightarrow 3 \):

由2成立以及定理9.3.9,有\( \lim_{x \to x_0; x \in X} f(x) = f(x_0) \),进而可得\( \forall \epsilon > 0, \exists \delta > 0, \forall x \in X, |x - x_0| < \delta, |f(x) - f(x_0)| \leq \dfrac{\epsilon}{2} < \epsilon \),即3成立。

\( 3 \Longrightarrow 4 \)是明显的,这里略过。

证明\( 4 \Longrightarrow 1 \):

\( \forall \epsilon > 0 \),由4可得,\( \exists \delta' > 0, \forall x \in X, |x - x_0| \leq \delta', |f(x) - f(x_0)| \leq \epsilon \),令\( \delta := 2\delta' \),可得\( \delta' < \delta \),于是有\( |x - x_0| < \delta \),综上,有\( \lim_{x \to x_0; x \in X} f(x) = f(x_0) \),进而有\( f \)在\( x_0 \)连续,即1成立。

最后根据传递性,有\( 1 \equiv 2 \equiv 3 \equiv 4 \)。

证毕。

练习9.4.2

题目:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( c \in \mathbf{R} \). Show that the constant function \( f: X \to \mathbf{R} \) defined by \( f(x) := c \) is continuous, and show that the identity function \( g: X \to \mathbf{R} \) defined by \( g(x) := x \) is also continuous.

证明:

证明f(x)连续:

\( \forall y \in X, \forall \epsilon > 0 \),令\( \delta := 1 \),有\( \forall x \in X, |x - y| < \delta, |f(x) - f(y)| = |c - c| = 0 < \epsilon \),根据定理9.4.7,有\( f \)在\( y \)连续。综上,有\( f \)连续。

证明g(x)连续:

\( \forall y \in X, \forall \epsilon > 0 \),令\( \delta := \dfrac{\epsilon}{2} \),有\( \forall x \in X, |x - y| < \delta, |g(x) - g(y)| = |x - y| < \delta = \dfrac{\epsilon}{2} < \epsilon \),根据定理9.4.7,有\( g \)在\( y \)连续。综上,有\( g \)连续。

证毕。

练习9.4.3

题目:

Prove Proposition 9.4.10. (Hint: you can use Lemma 6.5.3, combined with the squeeze test (Corollary 6.4.14) and Proposition 6.7.3.)

Proposition 9.4.10的内容:

(Exponentiation is continuous, I). Let \( a > 0 \) be a positive real number. Then the function \( f: \mathbf{R} \to \mathbf{R} \) defined by \( f(x) := a^x \) is continuous.

注:

Hint似乎是有问题的,这里不按Hint的思路去做。

思路:

考虑\( \forall y \in \mathbf{R} \),我们打算证明 \( \forall \epsilon > 0, \exists \delta > 0, \forall x \in \mathbf{R}, |x - y| \leq \delta, |f(x) - f(y)| = |a^x - a^y| \leq \epsilon \),然后再用定理9.4.7,这里反推下,提取出常数\( a^y \)(而不要提取变量\( a^x \)),可得\( |a^x - a^y| = |a^y||a^{x - y} - 1| \),因此\( |a^x - a^y| \leq \epsilon \) 相当于\( |a^y||a^{x - y} - 1| \leq \epsilon \),进而可得\( |a^{x - y} - 1| \leq \dfrac{\epsilon}{|a^y|} \),现在我们要找一个范围\( \delta > 0 \),使得\( \forall x \in \mathbf{R}, |x - y| \leq \delta \),均有\( |a^{x - y} - 1| \leq \dfrac{\epsilon}{|a^y|} \)成立,考虑到如果\( x \)和\( y \)足够近的话,即\( (x - y) \to 0 \),则\( a^{x - y} \to (a^0 = 1) \),此时\( |a^{x - y} - 1| \)也会变小,进而\( \leq \dfrac{\epsilon}{|a^y|} \),因此我们的目标转向让\( x, y \)足够靠近,可以通过两个数“夹着”以限制两者的范围,使它们足够靠近,具体细节见证明。

证明:

\( \forall y \in \mathbf{R}, \forall \epsilon > 0 \),由\( a > 0 \)以及引理6.5.3,有\( \lim_{n \to \infty} a^{1/n} = 1 \) (1) ,由定理6.7.3(或者说定理5.6.9的命题1),可得 \( \forall n \in \mathbf{N}, a^{1/n} \neq 0 \)以及\( a^y \neq 0 \),根据极限的倒数运算法则,可得\( \lim_{n \to \infty} a^{-1/n} = \lim_{n \to \infty} \dfrac{1}{a^{1/n}} = \dfrac{1}{\lim_{n \to \infty} a^{1/n}} = \dfrac{1}{1} = 1 \) (2) ,此时根据(1)和(2),可得\( \exists N_0 \in \mathbf{N}, \forall n \geq N_0, |a^{1/n} - 1| \leq \dfrac{\epsilon}{|a^y|} \) 以及\( \exists N_1 \in \mathbf{N}, \forall n \geq N_1, |a^{-1/n} - 1| \leq \dfrac{\epsilon}{|a^y|} \) (3) ,令\( N := \max(N_0, N_1, 1) \),令\( \delta := \dfrac{1}{N} \),则\( \forall x \in \mathbf{R}, |x - y| \leq \delta \),有\( -\delta \leq x - y \leq \delta \),根据定理6.7.3(或者说定理5.6.9的命题5),有\( a^{-\delta} \leq a^{x - y} \leq a^{\delta} \)或\( a^{\delta} \leq a^{x - y} \leq a^{-\delta} \) (4) (具体情况取决于\( x \geq 1 \)还是\( 0 < x < 1 \)),由\( N \geq N_0, N \geq N_1 \)以及(3)可得, \( a^{\delta}, a^{-\delta} \)均与\( 1 \) \( (\dfrac{\epsilon}{|a^y|}) \)-接近,根据定理4.3.7的命题6(对应的实数版本)以及(4),有\( a^{x - y} \)与\( 1 \)也\( (\dfrac{\epsilon}{|a^y|}) \)-接近,即\( |a^{x - y} - 1| \leq \dfrac{\epsilon}{|a^y|} \),进而可得\( |a^x - a^y| = |a^y||a^{x - y} - 1| \leq \epsilon \)。简而言之,\( \forall y \in \mathbf{R}, \forall \epsilon > 0 \), \( \exists \delta > 0, \forall x \in \mathbf{R}, |x - y| \leq \delta, |f(x) - f(y)| = |a^x - a^y| \leq \epsilon \),根据定理9.4.7,有\( f \)在\( y \)连续。综上,有\( f \)连续。

证毕。

练习9.4.4

题目:

Prove Proposition 9.4.11. (Hint: from limit laws (Proposition 9.3.14) one can show that \( \lim_{x \to 1} x^n = 1 \) for all integers \( n \). From this and the squeeze test (Corollary 6.4.14) deduce that \( \lim_{x \to 1} x^p = 1 \) for all real numbers \( p \). Finally, apply Proposition 6.7.3.)

Proposition 9.4.11的内容:

(Exponentiation is continuous, II). Let \( p \) be a real number. Then the function \( f: (0, +\infty) \to \mathbf{R} \) defined by \( f(x) := x^p \) is continuous.

注:

该练习如果按照Hint给的思路,则需要用到定理9.4.13,这是练习9.4.5证明的定理,不过练习9.4.5的证明不依赖于该练习的结果,因此不会产生循环依赖,可以使用。具体见作者博客的评论中Issa Rice、Friedrich以及作者的讨论, Issa Rice评论的时间是21 October, 2021 at 9:16 am,应该很好找到对应的评论,作者的勘误是将定理9.4.13移到定理9.4.10之前,对应将练习9.4.5移到练习9.4.3之前,定理9.4.13可以只移到定理9.4.11之前(练习也同理),但是定理9.4.10以及定理9.4.11是相关的,因此为了逻辑顺畅,还是移到定理9.4.10之前较好。

证明:

令\( a \)为任意非负实数,\( b \)为任意\( > a \)的正实数或者为\( +\infty \),同时要求\( 1 \)为\( (a, b) \)的附着点,可得\( \forall x \in (a, b), x > 0 \) (注意,开区间,取不到\( a \),所以\( x \neq 0 \)),下面用数学归纳法证明\( \forall n \in \mathbf{N}, \lim_{x \to 1; x \in (a, b)} x^n = 1 \):

当\( n = 0 \)时,\( \forall x \in (a, b), x^n = x^0 = 1 \),函数为常数函数,根据练习9.4.2,有\( \lim_{x \to 1; x \in (a, b)} x^n = 1 \),基础情况成立。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时,根据归纳假设,有\( \lim_{x \to 1; x \in (a, b)} x^k = 1 \),根据练习9.4.2,有\( \lim_{x \to 1; x \in (a, b)} x = 1 \),根据定理9.3.14,有\( \lim_{x \to 1; x \in (a, b)} x^{k + 1} = (\lim_{x \to 1; x \in (a, b)} x^k)(\lim_{x \to 1; x \in (a, b)} x) = 1 \times 1 = 1 \),即当\( n = k + 1 \)时也成立,归纳完毕。

\( \forall n \in \mathbf{Z} \)时,如果\( n \in \mathbf{N} \),则根据前面的证明,有\( \lim_{x \to 1; x \in (a, b)} x^n = 1 \),如果\( n < 0 \),则\( -n \in \mathbf{N} \),此时有\( \lim_{x \to 1; x \in (a, b)} x^{-n} = 1 \),根据定理6.7.3(或者说引理5.6.9的命题1),有\( \forall x \in (a, b), x^{-n} \neq 0 \),进而根据定理9.3.14,有\( \lim_{x \to 1; x \in (a, b)} x^n = \dfrac{\lim_{x \to 1; x \in (a, b)} 1}{\lim_{x \to 1; x \in (a, b)} x^{-n}} = \dfrac{1}{1} = 1 \)。综上,有\( \forall n \in \mathbf{Z}, \lim_{x \to 1; x \in (a, b)} x^n = 1 \)。

易证\( \exists n \in \mathbf{N} \)满足\( -n \leq p \leq n \):

  1. \( \forall x \in (1, +\infty) \),根据定理6.7.3(或者说引理5.6.9的命题5),有\( x^{-n} \leq x^p \leq x^n \),根据前面的证明,有\( \lim_{x \to 1; x \in (1, +\infty)} x^n = 1 = \lim_{x \to 1; x \in (1, +\infty)} x^{-n} \),根据练习9.3.5,有\( \lim_{x \to 1; x \in (1, +\infty)} x^p = 1 \)。
  2. \( \forall x \in (0, 1) \),根据定理6.7.3(或者说引理5.6.9的命题5),有\( x^n \leq x^p \leq x^{-n} \),根据前面的证明,有\( \lim_{x \to 1; x \in (0, 1)} x^n = 1 = \lim_{x \to 1; x \in (0, 1)} x^{-n} \),根据练习9.3.5,有\( \lim_{x \to 1; x \in (0, 1)} x^p = 1 \)。

我们现在证明\( \lim_{x \to 1; x \in (0, +\infty)} x^p = 1 \), \( \forall \epsilon > 0 \),因为\( \lim_{x \to 1; x \in (1, +\infty)} x^p = 1 \),因此\( \exists \delta_0 > 0, \forall x \in (1, +\infty), |x - 1| < \delta_0, |x^p - 1| \leq \epsilon \),因为\( \lim_{x \to 1; x \in (0, 1)} x^p = 1 \),因此\( \exists \delta_1 > 0, \forall x \in (0, 1), |x - 1| < \delta_1, |x^p - 1| \leq \epsilon \),取\( \delta = \min(\delta_0, \delta_1) \),有\( \forall x \in (1, +\infty), |x - 1| < \delta \leq \delta_0, |x^p - 1| \leq \epsilon \) 且\( \forall x \in (0, 1), |x - 1| < \delta \leq \delta_1, |x^p - 1| \leq \epsilon \),除此之外,当\( x = 1 \)时,\( |x - 1| = 0 < \delta, |x^p - 1| = |1 - 1| = 0 \leq \epsilon \),综合可得,\( \forall x \in (0, +\infty), |x - 1| < \delta, |x^p - 1| \leq \epsilon \),即\( \lim_{x \to 1; x \in (0, +\infty)} x^p = \lim_{x \to 1; x \in (0, +\infty)} f(x) = 1 = f(1) \),也就是\( f \)在\( 1 \)连续。

\( \forall y \in \mathbf(0, +\infty) \),我们要证明\( f(x) \)在\( y \)连续,构造函数\( g: (0, +\infty) \to (0, +\infty), \forall x \in (0, +\infty) \),令\( g(x) := \dfrac{x}{y} \),注意,对\( g \)来说,\( y \)是一个固定的常数,因为我们是针对每个\( y \)都构造一个对应的函数\( g \),而不是一个函数\( g \)对应所有\( y \),此时根据函数\( x \)在\( y \)连续、常数函数\( y \)在\( y \)连续以及定理9.4.9,有 \( g \)在\( y \)连续,又\( f \)在\( g(y) = \dfrac{y}{y} = 1 \)连续,根据定理9.4.13,有\( f \circ g \)在\( y \)连续,根据定理6.7.3(或者说引理5.6.9的命题2和命题3),有\( \forall x \in (0, +\infty), (\dfrac{x}{y})^p = \dfrac{x^p}{y^p} \),接着由\( f \circ g \)和常数函数\( y^p \)均在\( y \)连续以及定理9.4.9,有函数\( (f \circ g) \times y^p = (\dfrac{x^p}{y^p}) \times y^p = x^p \)在\( y \)连续,即\( f(x) \)在\( y \)连续。综上,有\( f \)连续。

证毕。

练习9.4.5

题目:

Prove Proposition 9.4.13.

Proposition 9.4.13的内容:

(Composition preserves continuity). Let \( X \) and \( Y \) be subsets of \( \mathbf{R} \), and let \( f: X \to Y \) and \( g: Y \to \mathbf{R} \) be functions. Let \( x_0 \) be a point in \( X \). If \(f \) is continuous at \( x_0 \), and \( g \) is continuous at \( f(x_0) \), then the composition \( g \circ f: X \to \mathbf{R} \) is continuous at \( x_0 \).

证明:

因为\( f \)在\( x_0 \)连续,因此\( \lim_{x \to x_0; x \in X} f(x) = f(x_0) \) (1) ,因为\( g \)在\( f(x_0) \)连续,因此\( \lim_{y \to f(x_0); y \in Y} g(y) = g(f(x_0)) \) (2) , \( \forall \epsilon > 0 \),由(2)以及定理9.4.7可得,\( \exists \delta_0 > 0, \forall y \in Y, |y - f(x_0)| \leq \delta_0, |g(y) - g(f(x_0))| \leq \epsilon \) (3) ,进而由(1)以及定理9.4.7可得,\( \exists \delta > 0, \forall x \in X, |x - x_0| \leq \delta, |f(x) - f(x_0)| \leq \delta_0 \),再由\( f(x) \in Y \)以及(3)可得,\( |g(f(x)) - g(f(x_0))| \leq \epsilon \)。综上,有\( \forall \epsilon > 0, \exists \delta > 0, \forall x \in X, |x - x_0| \leq \delta, |(g \circ f)(x) - (g \circ f)(x_0| = |g(f(x)) - g(f(x_0))| \leq \epsilon \),再由定理9.4.7,可得\( g \circ f \)在\( x_0 \)连续。

证毕。

练习9.4.6

题目:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( f: X \to \mathbf{R} \) be a continuous function. If \( Y \) is a subset of \( X \), show that the restriction \( f|_Y : Y \to \mathbf{R} \) of \( f \) to \( Y \) is also a continuous function. (Hint: this is a simple result, but it requires you to follow the definitions carefully.)

证明:

\( \forall y \in Y \),因为\( Y \subseteq X \),因此\( y \in X \),再由\( f \)在\( y \)连续以及定理9.4.7,可得\( \forall \epsilon > 0, \exists \delta > 0, \forall x \in X, |x - y| \leq \delta, |f(x) - f(y)| \leq \epsilon \) (1) ,而\( \forall x \in Y, |x - y| \leq \delta \),因为\( Y \subseteq X \),有\( x \in X, |x - y| \leq \delta \),进而根据(1),有\( |f(x) - f(y)| \leq \epsilon \),而\( f|_Y(x) = f(x), f|_Y(y) = f(y) \)(因为\( x, y \in Y \)),因此也有\( |f|_Y(x) - f|_Y(y)| \leq \epsilon \),简而言之,\( \forall \epsilon > 0, \exists \delta > 0, \forall x \in Y, |x - y| \leq \delta, |f|_Y(x) - f|_Y(y)| \leq \epsilon \),再次根据定理9.4.7,有\( f|_Y \)在\( y \)连续。综上,有\( f|_Y \)连续。

证毕。

练习9.4.7

题目:

Let \( n \geq 0 \) be an integer, and for each \( 0 \leq i \leq n \) let \( c_i \) be a real number. Let \( P: \mathbf{R} \to \mathbf{R} \) be the function \( P(x) := \sum_{i = 0}^{n} c_ix^i \); such a function is known as a polynomial of one variable; a typical example is \( P(x) = 6x^4 - 3x^2 + 4 \). Show that \( P \) is continuous.

注:

定理9.4.11中函数\( f \)的定义域是\( (0, +\infty) \),不是\( \mathbf{R} \),在此练习的证明中不能直接用,这里直接不用。

证明:

\( n \)为非负整数,即自然数。

\( \forall c \in \mathbf{R}, \forall n \in \mathbf{N} \),定义函数\( g: \mathbf{R} \to \mathbf{R}, \forall x \in \mathbf{R}, g(x) := cx^n \),我们证明\( g \)连续 (1) ,用数学归纳法证明:

当\( n = 0 \)时,\( \forall x \in \mathbf{R}, cx^n = cx^0 = c \),根据练习9.4.2,有\( g \)连续,基础情况成立。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时,根据归纳假设,有函数\( cx^k \)连续,根据练习9.4.2,有函数\( x \)连续,于是\( \forall y \in \mathbf{R} \),有函数\( cx^k \)在\( y \)连续,函数\( x \)在\( y \)连续,进而根据定理9.4.9,有\( g(x) = cx^{k + 1} = c(x^kx) = (cx^k)x \)在\( y \)连续。综上,有\( g \)连续,即当\( n = k + 1 \)时也成立,归纳完毕。

最后回去证明题目,用数学归纳法证明:

当\( n = 0 \)时,\( P(x) := c_0x^0 \),根据(1),有\( P \)连续,基础情况成立。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时, \( P(x) := \sum_{i = 0}^{k + 1} c_ix^i = (\sum_{i = 0}^{k} c_ix^i) + c_{k + 1}x^{k + 1} \),根据归纳假设,\( \sum_{i = 0}^{k} c_ix^i \)连续,根据(1),有\( c_{k + 1}x^{k + 1} \)连续,于是\( \forall y \in \mathbf{R} \),有\( \sum_{i = 0}^{k} c_ix^i \)在\( y \)连续,以及\( c_{k + 1}x^{k + 1} \)在\( y \)连续,进而根据定理9.4.9,有\( P(x) = (\sum_{i = 0}^{k} c_ix^i) + c_{k + 1}x^{k + 1} \)在\( y \)连续。综上,有\( P \)连续,即\( n = k + 1 \)也成立,归纳完毕。

证毕。

章节9.5

练习9.5.1

题目:

Let \( E \) be a subset of \( \mathbf{R} \), let \( f: E \to \mathbf{R} \) be a function, and let \( x_0 \) be an adherent point of \( E \). Write down a definition of what it would mean for the limit \( \lim_{x \to x_0; x \in E} f(x) \) to exist and equal \( +\infty \) or \( -\infty \). If \( f : \mathbf{R} \setminus \{ 0 \} \to \mathbf{R} \) is the function \( f(x):= 1/x \), use your definition to conclude \( f(0+) = +\infty \) and \( f(0−) = −\infty \). Also, state and prove some analogue of Proposition 9.3.9 when \( L = +\infty \) or \( L = -\infty \).

思路:

关键是在\( x \to x_0 \)的过程中,函数值的“趋势”保持一致,一起走向\( +\infty \) 或者一起走向\( -\infty \),而不要“上蹿下跳”。

解答:

定义\( \lim_{x \to x_0; x \in E} f(x) \)等于\( +\infty \)或\( -\infty \):

定义\( \lim_{x \to x_0; x \in E} f(x) := +\infty \)当且仅当 \( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists \delta > 0, \forall x \in E, |x - x_0| < \delta, f(x) > N \)。定义\( \lim_{x \to x_0; x \in E} f(x) := -\infty \)当且仅当 \( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists \delta > 0, \forall x \in E, |x - x_0| < \delta, f(x) < -N \)。

证明\( f(0+) = +\infty \):

\( \forall N \in \mathbf{N} \setminus \{ 0 \} \),有\( \exists x' > 0 \)满足\( x' < \dfrac{1}{N} \),从而有\( \dfrac{1}{x'} > N \),令\( \delta = x' \),令\( E := (\mathbf{R} \setminus \{ 0 \}) \cap (0, +\infty) \),则\( 0 \)是\( E \)的附着点且\( \forall x \in E, |x - 0| < \delta \),有\( |x| = x < \delta = x' \),即\( x < x' \),进而\( f(x) = \dfrac{1}{x} > \dfrac{1}{x'} > N \)。简而言之,\( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists \delta > 0, \forall x \in E, |x - 0| < \delta, f(x) > N \),也就是\( \lim_{x \to 0; x \in E} = \lim_{x \to 0; x \in (\mathbf{R} \setminus \{ 0 \}) \cap (0, +\infty)} = +\infty \),再根据左极限的定义,有\( f(0+) = +\infty \)。

证毕。

证明\( f(0-) = -\infty \):

\( \forall N \in \mathbf{N} \setminus \{ 0 \} \),有\( \exists x' < 0 \)满足\( 0 < -x' < \dfrac{1}{N} \),进而\( -\dfrac{1}{N} < x' < 0 \),从而有\( \dfrac{1}{x'} < -N < 0 \),令\( \delta = -x' \),令\( E := (\mathbf{R} \setminus \{ 0 \}) \cap (-\infty, 0) \),则\( 0 \)是\( E \)的附着点且\( \forall x \in E, |x - 0| < \delta \),有\( |x| = -x < \delta = -x' \),即\( -x < -x' \),也就是\( x' < x < 0 \),进而\( f(x) = \dfrac{1}{x} < \dfrac{1}{x'} < -N \)。简而言之,\( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists \delta > 0, \forall x \in E, |x - 0| < \delta, f(x) < -N \),也就是\( \lim_{x \to 0; x \in E} = \lim_{x \to 0; x \in (\mathbf{R} \setminus \{ 0 \}) \cap (-\infty, 0)} = -\infty \),再根据左极限的定义,有\( f(0-) = -\infty \)。

证毕。

定理9.3.9在\( L = +\infty \)时的版本:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( L = +\infty \). Then the following two statements are logically equivalent:

  1. \( \lim_{x \to x_0; x \in E} f(x) = L \)
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), we have \( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists K \in \mathbf{N}, \forall n \geq K, f(a_n) > N \).

证明:

必要性:

\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in E \)且\( \lim_{n \to \infty} a_n = x_0 \),由\( \lim_{x \to x_0; x \in E} f(x) = L = +\infty \),可得 \( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists \delta > 0, \forall x \in E, |x - x_0| < \delta, f(x) > N \),因为\( \lim_{n \to \infty} a_n = x_0 \),因此\( \exists K \in \mathbf{N}, \forall n \geq K, |a_n - x_0| \leq \dfrac{\delta}{2} < \delta \),进而有\( f(a_n) > N \)。简而言之,\( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists K \in \mathbf{N}, \forall n \geq K, f(a_n) > N \),即2成立。

充分性:

假设\( \lim_{x \to x_0; x \in E} f(x) = L \)不成立,即 \( \exists N_0 \in \mathbf{N} \setminus \{ 0 \}, \forall \delta > 0, \exists x \in E, |x - x_0| < \delta, f(x) \leq N_0 \) (1) ,此时\( \forall n \in \mathbf{N} \),令\( \delta_n := \dfrac{1}{n + 1} \),根据(1),有\( \exists x_n \in E, |x_n - x_0| < \delta_n, f(x_n) \leq N_0 \),则令\( X_n := \{ f(x) \leq N_0 : x \in E, |x - x_0| < \delta_n \} \),有\( X_n \)非空,使用选择公理,可以得到序列\( (a_n)_{n = 0}^{\infty} \)满足\( \forall n \in \mathbf{N}, a_n \in X_n \),特别的,有\( a_n \in E \),易证\( \lim_{n \to \infty} a_n = x_0 \),然而由\( a_n \in X_n \)可得, \( \forall n \in \mathbf{N} \),均有\( f(a_n) \leq N_0 \),可得2不成立,矛盾,因此假设不成立,有\( \lim_{x \to x_0; x \in E} f(x) = L \)。

证毕。

定理9.3.9在\( L = -\infty \)时的版本:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a function, let \( E \) be a subset of \( X \), let \( x_0 \) be an adherent point of \( E \), and let \( L = -\infty \). Then the following two statements are logically equivalent:

  1. \( \lim_{x \to x_0; x \in E} f(x) = L \)
  2. For every sequence \( (a_n)_{n = 0}^{\infty} \) which consists entirely of elements of \( E \) and converges to \( x_0 \), we have \( \forall N \in \mathbf{N} \setminus \{ 0 \}, \exists K \in \mathbf{N}, \forall n \geq K, f(a_n) < -N \).

由于对称性,\( L = -\infty \)时的版本的证明和\( L = +\infty \)基本一致,这里省略\( L = -\infty \)时的版本的证明。

章节9.6

练习9.6.1

题目:

Give examples of

  1. a function \( f: (1, 2) \to \mathbf{R} \) which is continuous and bounded, attains its minimum somewhere, but does not attain its maximum anywhere;
  2. a function \( f: [0, +\infty) \to \mathbf{R} \) which is continuous, bounded, attains its maximum somewhere, but does not attain its minimum anywhere;
  3. a function \( f: [-1, 1] \to \mathbf{R} \) which is bounded but does not attain its minimum anywhere or its maximum anywhere.
  4. a function \( f: [−1, 1] \to \mathbf{R} \) which has no upper bound and no lower bound.

Explain why none of the examples you construct violate the maximum principle. (Note: read the assumptions carefully!)

解答:

1的例子:

定义函数\( f: (1, 2) \to \mathbf{R}, \forall x \in (1, 2) \),如果\( x \in (0, 0.5] \),则令\( f(x) := 0.5 \),否则,令\( f(x) := x \)。 \( \forall x \in (1, 2) \),有\( |f(x)| \leq 3 \),即\( f \)有界,易证\( f \)连续(\( 0.5 \)似乎是不连续点,但是可以证明\( 0.5 \)这点的左右极限都是\( f(0.5) = 0.5 \),再用定理9.5.3,就能证明\( f \)在\( 0.5 \)连续了),且\( f \)明显在\( 0.5 \)达到最小值\( f(0.5) = 0.5 \),接下来我们证明\( f \)达不到最大值,假设\( f \)能达到最大值,即\( \exists x_0 \in (1, 2), \forall x \in (1, 2), f(x) \leq f(x_0) \) (1)

  1. 如果\( x_0 \in (0, 0.5] \),则\( f(x_0) = 0.5 < f(0.6) \),这和(1)矛盾。
  2. 如果\( x_0 \in (0.5, 2) \),此时有\( f(x_0) = x_0 \),但\( x_0 + \dfrac{2 - x_0}{2} \in (1, 2) \)且 \( f(x_0 + \dfrac{2 - x_0}{2}) = x_0 + \dfrac{2 - x_0}{2} > f(x_0) \),这和(1)矛盾。

综上,所有情况都矛盾,因此假设不成立,有\( f \)达不到最大值。

2的例子:

定义函数\( f: [0, +\infty) \to \mathbf{R}, \forall x \in [0, +\infty) \),如果\( x \in [0, 1] \),则令\( f(x) := 1 \),否则,令\( f(x) := \dfrac{1}{x} \)。易证\( f \)连续(易证\( 1 \)处左右极限都是\( f(1) = 1 \),再用定理9.5.3就行),也易证\( f \)有界(因为\( 0 < f(x) \leq 1 \),因此\( |f(x)| \leq 1 \)),且\( f \)在\( 1 \)达到最大值\( f(1) = 1 \),接下来我们证明\( f \)达不到最小值,假设\( f \)能达到最小值,即\( \exists x_0 \in [0, +\infty), \forall x \in [0, +\infty), f(x_0) \leq f(x) \) (1)

  1. 如果\( x_0 \in [0, 1] \),则\( f(x_0) = 1 > f(2) \),这和(1)矛盾。
  2. 如果\( x_0 \in (1, +\infty) \),此时有\( f(x_0) = \dfrac{1}{x_0} \),但\( x_0 + 1 \in [0, +\infty) \)且 \( f(x_0 + 1) = \dfrac{1}{x_0 + 1} < f(x_0) \),这和(1)矛盾。

综上,所有情况都矛盾,因此假设不成立,有\( f \)达不到最小值。

3的例子:

定义函数\( f: [-1, 1] \to \mathbf{R}, \forall x \in [-1, 1] \),如果\( x \in (-0.5, 0.5) \),则令\( f(x) := x \),否则,令\( f(x) := 0 \)。明显\( f \)是有界的。

先证明\( f \)达不到最大值,假设\( f \)能达到最大值,即 \( \exists x_0 \in [-1, 1], \forall x \in X, f(x) \leq f(x_0) \) (1) :明显有

  1. 如果\( x_0 \in (-0.5, 0.5) \),则\( f(x_0) = x_0 \),此时令\( x_1 := x_0 + \dfrac{0.5 - x_0}{2} \),有\( x_1 \in (-0.5, 0.5) \) 且\( f(x_1) = x_1 > f(x_0) \),这和(1)矛盾。
  2. 如果\( x_0 \in [-1, -0.5] \cup [0.5, 1] \),则\( f(x_0) = 0 < 0.1 = f(0.1) \),这和(1)矛盾。

综上,所有情况都矛盾,因此假设不成立,有\( f \)达不到最大值。

最后证明\( f \)达不到最小值,假设\( f \)能达到最小值,即 \( \exists x_0 \in [-1, 1], \forall x \in X, f(x_0) \leq f(x) \) (2) :明显有

  1. 如果\( x_0 \in (-0.5, 0.5) \),则\( f(x_0) = x_0 \),此时令\( x_1 := x_0 + \dfrac{-0.5 - x_0}{2} \),有\( x_1 \in (-0.5, 0.5) \) 且\( f(x_1) = x_1 < f(x_0) \),这和(2)矛盾。
  2. 如果\( x_0 \in [-1, -0.5] \cup [0.5, 1] \),则\( f(x_0) = 0 > -0.1 = f(-0.1) \),这和(2)矛盾。

综上,所有情况都矛盾,因此假设不成立,有\( f \)达不到最小值。

4的例子:

定义函数\( f: [-1, 1] \to \mathbf{R}, \forall x \in [-1, 1] \),如果\( x \neq 0 \),则令\( f(x) := \dfrac{1}{x} \),否则令\( f(x) := 0 \),易证\( x \)从左边靠近\( 0 \), \( f(x) \)会趋向于\( -\infty \),因此\( f \)无下界,同理,\( x \)从右边靠近\( 0 \), \( f(x) \)会趋向于\( +\infty \),因此\( f \)无上界。

解释所有构造的例子为什么都没有违反最大值原理:

例子1和例子2的区间不是闭区间,不满足最大值原理的前提条件,因此不违反最大值原理。

例子3和例子4不是指定区间上的连续函数,不满足最大值原理的前提条件,因此不违反最大值原理。

练习9.6.2

题目:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f, g: X \to \mathbf{R} \) be bounded functions, show that \( f + g, f - g \), and \( f \cdot g \) are also bounded functions. If we furthermore assume that \( g(x) \neq 0 \) for all \( x \in X \), is it true that \( f / g \) is bounded? Prove this or give a counterexample.

注:

这是作者博客的勘误中增加的练习。

证明\( f + g, f - g, f \cdot g \)有界:

因为\( f \)有界,因此\( \exists M_0 \in \mathbf{R}, \forall x \in X, |f(x)| \leq M_0 \),因为\( g \)有界,因此\( \exists M_1 \in \mathbf{R}, \forall x \in X, |g(x)| \leq M_1 \)。 \( \forall x \in X \),有\( |(f + g)(x)| = |f(x) + g(x)| \leq |f(x)| + |g(x)| \leq M_0 + M_1 \),即\( f + g \)有界。 \( \forall x \in X \),有\( |(f - g)(x)| = |f(x) - g(x)| \leq |f(x)| + |-g(x)| = |f(x)| + |g(x)| \leq M_0 + M_1 \),即\( f - g \)有界。 \( \forall x \in X \),有\( |(f \cdot g)(x)| = |f(x)g(x)| = |f(x)||g(x)| \leq M_0M_1 \),即\( f \cdot g \)有界。

证毕。

解答\( f / g \)一定有界是否成立:

\( f / g \)一定有界不成立,举个例子,令\( X := (0, 1) \),令\( f(x) := 1 \),令\( g(x) := x \),这两个函数都是有界的,且\( \forall x \in X, g(x) \neq 0 \),但\( x \)趋向于\( 0 \)时, \( (f / g)(x) = \dfrac{1}{x} \)会趋向于正无穷,因此无界。

章节9.7

练习9.7.1

题目:

Prove Corollary 9.7.4. (Hint: you may need Exercise 9.4.6 in addition to the intermediate value theorem.)

Corollary 9.7.4的内容:

(Images of continuous functions). Let \( a < b \), and let \( f: [a, b] \to \mathbf{R} \) be a continuous function on \( [a, b] \). Let \( M := \sup_{x \in [a, b]} f(x) \) be the maximum value of \( f \), and let \( m := \inf_{x \in [a, b]} f(x) \) be the minimum value. Let \( y \) be a real number between \( m \) and \( M \) (i.e., \( m \leq y \leq M \) ). Then there exists a \( c \in [a, b] \) such that \( f( c) = y \). Furthermore, we have \( f([a, b]) = [m, M] \).

证明:

根据定理9.6.7,\( \exists x_M \in [a, b], f(x_M) = M, \exists x_m \in [a, b], f(x_m) = m \)。如果\( x_M = x_m \),则\( m = y = M \),此时令\( c := x_M \),有\( f( c) = y \),满足要求。如果\( x_M \neq x_m \),此时不妨设\( x_m < X_M \),因为\( x_m \in [a, b], x_M \in [a, b] \),因此\( [x_m, x_M] \subseteq [a, b] \),令\( g := f|_{[x_m, x_M]} \),根据练习9.4.6,有\( g \)也连续,而\( g(x_m) = f(x_m) = m \leq y \leq M = f(x_M) = g(x_M) \),对\( g \)使用中值定理,有\( \exists c \in [x_m, x_M], f( c) = y \)。

最后证明下\( f([a, b]) = [m, M] \):先证明\( f([a, b]) \subseteq [m, M] \), \( \forall y \in f([a, b]) \),有\( \exists x \in [a, b], f(x) = y \),而\( m, M \)分别是\( f \)的最大、最小值,因此\( m \leq (f(x) = y) \leq M \),进而\( y \in [m, M] \)。接着,证明\( [m, M] \subseteq f([a, b]) \), \( \forall y \in [m, M] \),根据前面的证明,有\( \exists c \in [a, b], f( c) = y \),于是有\( (y = f( c)) \in f([a, b]) \)。综上,\( f([a, b]) \subseteq [m, M], [m, M] \subseteq f([a, b]) \),有\( f([a, b]) = [m, M] \)。

证毕。

练习9.7.2

题目:

Let \( f: [0, 1] \to [0, 1] \) be a continuous function. Show that there exists a real number \( x \) in \( [0, 1] \) such that \( f(x) = x \). (Hint: apply the intermediate value theorem to the function \( f(x) - x \).) This point \( x \) is known as a fixed point of \( f \), and this result is a basic example of a fixed point theorem, which play an important role in certain types of analysis.

证明:

定义函数\( f': [0, 1] \to \mathbf{R}, \forall x \in [0, 1] \),令\( f'(x) := f'(x) \),易证\( f' \)连续。定义函数\( g: [0, 1] \to \mathbf{R}, \forall x \in [0, 1] \),令\( g(x) := f'(x) - x \),根据定理9.4.9,可得\( g \)也连续。\( f \)的值域范围是\( [0, 1] \),因此\( f'([0, 1]) \subseteq [0, 1] \),特别的,有\( f'(1) \leq 1 \),进而\( g(1) = f'(1) - 1 \leq 0 \),同理,有\( g(0) = f'(0) - 0 \geq 0 \),因此\( g(1) \leq 0 \leq g(0) \),根据中值定理,有\( \exists x \in [0, 1], g(x) = 0 \),也就是\( f'(x) = x \),进而可得\( f(x) = x \)。

证毕。

章节9.8

练习9.8.1

题目:

Explain why the maximum principle remains true if the hypothesis that \( f \) is continuous is replaced with \( f \) being monotone, or with \( f \) being strictly monotone. (You can use the same explanation for both cases.)

解释:

令\( a < b \)为实数,\( f: [a, b] \to \mathbf{R} \)为\( [a, b] \)上的单调递增函数,此时\( \forall x \in [a, b] \),有\( a \leq x \leq b \),因此\( f(a) \leq f(x) \leq f(b) \),可得\( f \)在\( b \)达到最大值,在\( a \)达到最小值。

严格单调递增函数、单调递减函数、严格单调递减函数同理,不再赘述。

练习9.8.2

题目:

Give an example to show that the intermediate value theorem becomes false if the hypothesis that \( f \) is continuous is replaced with \( f \) being monotone, or with \( f \) being strictly monotone. (You can use the same counterexample for both cases.)

解答:

定义函数\( f: [0, 1] \to \mathbf{R} \),\( \forall x \in [0, 1] \),如果\( 0 \leq x < 0.5 \),则令\( f(x) := x \),如果\( 0.5 < x \leq 1 \),则令\( f(x) := 2x \),如果\( x = 0.5 \),则令\( f(x) := 0.5 + (2 \times 0.5 - 0.5) / 2 = 0.5 + 0.25 = 0.75 \),明显\( f \)是严格单调递增的,但取\( y := 0.85 \),有\( f(0) = 0 \leq y \leq 2 = f(1) \),但\( \forall x \in [0, 1] \),均\( f(x) \neq y \)(分\( x \in [0, 0.5), x = 0.5, x \in (0.5, 1] \) 讨论下就行),可得中值定理不成立。

单调递增函数、单调递减函数、严格单调递减函数同理,不再赘述。

练习9.8.3

题目:

Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a function which is both continuous and one-to-one. Show that \( f \) is strictly monotone. (Hint: divide into the three cases \( f(a) < f(b) , f(a) = f(b), f(a) > f(b) \). The second case leads directly to a contradiction. In the first case, use contradiction and the intermediate value theorem to show that \( f \) is strictly monotone increasing; in the third case, argue similarly to show \( f \) is strictly monotone decreasing.)

证明:

因为\( f \)单射,因此\( f(a) \neq f(b) \)。

如果\( f(a) < f(b) \),此时我们先证明\( \forall x \in (a, b], f(a) < f(x) \),假设\( \exists x_0 \in (a, b], f(a) \geq f(x_0) \) (1) ,由\( f \)单射,可得\( f(a) > f(x_0) \),此时有\( f(x_0) < f(a) < f(b) \),我们想对\( f|_{[x_0, b]} \)使用中值定理,但是这还得保证\( x_0 < b \),此时已经有\( x_0 \leq b \)了(因为\( x_0 \in (a, b] \)),假设\( x_0 = b \),则\( f(x_0) = f(b) > f(a) \),这和(1)矛盾,因此\( x_0 \neq b \),有\( x_0 < b \),现在可以使用中值定理了,对\( f|_{[x_0, b]} \)使用中值定理,可得\( \exists x_1 \in [x_0, b], f(x_1) = f(a) \),因为\( x_0 > a \)(这是因为\( x_0 \in (a, b] \)),因此\( x_1 \in [x_0, b] \neq a \),但\( f(x_1) = f(a) \),这和\( f \)单射矛盾,因此假设不成立,有\( \forall x \in (a, b], f(a) < f(x) \)。我们现在证明\( f \)严格单调递增,假设\( f \)非严格单调递增,即\( \exists x_2 < x_3 \in [a, b], f(x_3) \leq f(x_2) \),又\( f \)单射,因此\( f(x_3) < f(x_2) \),明显有\( x_3 \neq a \)(因为\( x_3 \)左边还有个\( x_2 \)),因此\( x_3 \in (a, b] \),根据前面的证明,有\( f(a) < f(x_3) \),此时有\( f(a) < f(x_3) < f(x_2) \) (2) ,我们想对\( f|_{[a, x_2]} \)使用中值定理,但是这还得保证\( a < x_2 \),此时已经有\( a \leq x_2 \)了(因为\( x_2 \in [a, b]\)),假设\( a = x_2 \),则\( f(x_2) = f(a) < f(x_3) \),这和(2)矛盾,因此\( a \neq x_2 \),有\( a < x_2 \),对\( f|_{[a, x_2]} \)使用中值定理,可得\( \exists x_4 \in [a, x_2], f(x_4) = f(x_3) \),因为\( x_3 > x_2 \),因此\( x_4 \in [a, x_2] \neq x_3 \),但\( f(x_4) = f(x_3) \),这和\( f \)单射矛盾,因此假设不成立,有\( f \)严格单调递增。

同理易证在\( f(a) > f(b) \)时,\( f \)严格单调递减。

综上,\( f \)为严格单调函数。

证毕。

练习9.8.4

题目:

Prove Proposition 9.8.3. (Hint: to show that \( f^{-1} \) is continuous, it is easiest to use the “epsilon-delta” definition of continuity, Proposition 9.4.7(3).) Is the proposition still true if the continuity assumption is dropped, or if strict monotonicity is replaced just by monotonicity? How should one modify the proposition to deal with strictly monotone decreasing functions instead of strictly monotone increasing functions?

Proposition 9.8.3的内容:

Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a function which is both continuous and strictly monotone increasing. Then \( f \) is a bijection from \( [a, b] \) to \( [f(a), f(b)] \), and the inverse \( f^{−1}: [f(a), f(b)] \to [a, b] \) is also continuous and strictly monotone increasing.

证明\( f \)是\( [a, b] \to [f(a), f(b)] \)的双射函数:

将\( f \)的值域限定到\( [f(a), f(b)] \)。

先证明\( f \)单射,\( \forall x_1 \neq x_2 \in [a, b] \),有\( x_1 < x_2 \)或\( x_2 < x_1 \),不妨设\( x_1 < x_2 \),此时因为\( f \)严格单调递增,因此\( f(x_1) < f(x_2) \),这意味着\( f(x_1) \neq f(x_2) \)。综上,\( f \)单射。

接着证明\( f \)满射,\( \forall y \in [f(a), f(b)] \),有\( f(a) \leq y \leq f(b) \),根据中值定理,可得\( \exists x \in [a, b], f(x) = y \),综上,\( f \)满射。

综上,\( f \)为\( [a, b] \to [f(a), f(b)] \)的双射函数。

证毕。

证明\( f^{-1}: [f(a), f(b)] \to [a, b] \)也连续且严格单调递增:

先证明\( f^{-1} \)严格单调递增,\( \forall y_1 < y_2 \in [f(a), f(b)] \),因为\( f \)双射,因此\( \exists x_1, x_2 \in [a, b], f(x_1) = y_1, f(x_2) = y_2 \),可得\( f^{-1}(y_1) = x_1, f^{-1}(y_2) = x_2 \),假设\( x_1 \geq x_2 \),则由\( f \)严格单调递增,可得\( y_1 = f(x_1) \geq f(x_2) = y_2 \),这和\( y_1 < y_2 \)矛盾,因此假设不成立,有\( x_1 < x_2 \),即\( f^{-1}(y_1) < f^{-1}(y_2) \)。综上,\( f^{-1} \)严格单调递增。

接着证明\( f^{-1} \)连续,先证明\( f^{-1} \)在\( (f(a), f(b)) \)上连续,暂时不考虑端点, \( \forall y_t \in (f(a), f(b)) \),\( \forall \epsilon > 0 \),令\( x_t := f^{-1}(y_t) \),因为\( f^{-1} \)严格单调递增,因此\( b = f^{-1}(f(b)) > x_t, x_t > f^{-1}(f(a)) = a \),令\( \epsilon' := \min(\dfrac{b - x_t}{2}, \dfrac{x_t - a}{2}, \epsilon) \),可得\( 0 < \epsilon' \leq \epsilon, x_t + \epsilon' \in (a, b), x_t - \epsilon' \in (a, b) \),令\( y_{t+} := f(x_t + \epsilon'), y_{t-} := f(x_t - \epsilon') \),令\( \delta := \min(y_{t+} - y_t, y_t - y_{t-}) \),因为\( f \)严格单调递增,因此\( y_{t+} - y_t > 0, y_t - y_{t-} > 0 \),这意味着\( \delta > 0 \),此时\( \forall y \in [f(a), f(b)], |y - y_t| < \delta \),有\( y_{t-} < y_t - \delta < y < y_t + \delta < y_{t+} \),进而由\( f^{-1} \)严格单调递增,可得\( f^{-1}(y_{t-}) < f^{-1}(y) < f^{-1}(y_{t+}) \),即\( x_t - \epsilon' < f^{-1}(y) < x_t + \epsilon' \),可得\( |f^{-1}(y) - f^{-1}(y_t)| < \epsilon' \leq \epsilon \)(注:\( x_t = f^{-1}(y_t) \))。简而言之,\( \forall y_t \in (f(a), f(b)), \forall \epsilon > 0, \exists \delta > 0, \forall y \in [f(a), f(b)], |y - y_t| < \delta, |f^{-1}(y) - f^{-1}(y_t)| < \epsilon \),根据定理9.4.7的3,可得\( f^{-1} \)在\( (f(a), f(b)) \)上连续,还差证明在两个端点处连续。

接着证明\( f^{-1} \)在\( f(a) \)连续,\( \forall \epsilon > 0 \),令\( \epsilon' := \min(b - a, \epsilon) \),可得\( 0 < \epsilon' \leq \epsilon, a + \epsilon' \in [a, b] \),令\( y_{a+} := f(a + \epsilon') \),令\( \delta := y_{a+} - f(a) \),由\( f \)严格单调递增,可得\( y_{a+} > f(a) \),因此\( \delta > 0 \),此时\( \forall y \in [f(a), f(b)], |y - f(a)| < \delta \),有\( f(a) \leq y < f(a) + \delta < y_{a+} \),再由\( f^{-1} \)严格单调递增,可得\( a - \epsilon' < a \leq f^{-1}(y) < a + \epsilon' \),进而可得\( |f^{-1}(y) - a| < \epsilon' \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists \delta > 0, \forall y \in [f(a), f(b)], |y - a| < \delta, |f^{-1}(y) - f^{-1}(y_t)| < \epsilon \),根据定理9.4.7的3,可得\( f^{-1} \)在\( f(a) \)连续。

同理易证\( f^{-1} \)在\( f(b) \)连续。

综上,\( f^{-1} \)在\( [f(a), f(b)] \)上连续,即\( f^{-1} \)连续。

证毕。

如果将\( f \)连续的假设去掉,命题还成立吗:

不成立,举个例子,令\( a < b \)为实数,定义函数\( f: [a, b] \to \mathbf{R} \), \( \forall x \in [a, b] \),如果\( x \neq b \),则令\( f(x) := x \),如果\( x = b \),则令\( f(x) := b + 100 \)。此时\( [f(a), f(b)] = [a, b + 100] \),有\( f: [a, b] \to [f(a), f(b)] \)非双射,具体的,它不满射,因为\( b + 99 \in [f(a), f(b)] \),但不存在\( x \in [a, b] \),使得\( f(x) := b + 99 \),由于\( f: [a, b] \to [f(a), f(b)] \)非双射,自然不存在逆函数\( f^{-1}: [f(a), f(b)] \to [a, b] \),因此逆函数相关的结论也是无效的。

如果将“严格单调递增”改成“单调递增”,命题还成立吗:

注意,这里是将所有“严格单调递增”都改成“单调递增”,而不只是针对\( f \)修改。

不成立,主要问题是此时可能有两个点对应的函数值相等,即\( f \)非单射,举个例子,令\( a < b \)为实数,定义函数\( f: [a, b] \to \mathbf{R} \), \( \forall x \in [a, b] \),令\( f(x) := 1 \),此时\( f \)连续且单调递增,但\( f \)非单射,进而非双射,自然也不存在逆函数\( f^{-1}: [f(a), f(b)] \to [a, b] \)。

该命题对应的严格单调递减版本:

只要将“严格单调递增”改成“严格单调递减”,以及\( [f(a), f(b)] \)改成 \( [f(b), f(a)] \)(因为此时\( f(b) < f(a) \))就行,如下:

Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a function which is both continuous and strictly monotone decreasing. Then \( f \) is a bijection from \( [a, b] \) to \( [f(b), f(a)] \), and the inverse \( f^{−1}: [f(b), f(a)] \to [a, b] \) is also continuous and strictly monotone decreasing.

练习9.8.5

题目:

In this exercise we give an example of a function which has a discontinuity at every rational point, but is continuous at every irrational. Since the rationals are countable, we can write them as \( \mathbf{Q} = \{ q(0), q(1), q(2), \dots \} \), where \( q: \mathbf{N} \to \mathbf{Q} \) is a bijection from \( \mathbf{N} \) to \( \mathbf{Q} \). Now define a function \( g: \mathbf{Q} \to \mathbf{R} \) by setting \( g(q(n)) := 2^{-n} \) for each natural number \( n \); thus \( g \) maps \( q(0) \) to \( 1 \), \( q(1) \) to \( 2^{-1} \), etc. Since \( \sum_{n = 0}^{\infty} 2^{-n} \) is absolutely convergent, we see that \( \sum_{r \in Q} g( r) \) is also absolutely convergent. Now define the function \( f: \mathbf{R} \to \mathbf{R} \) by \( f(x) := \sum_{r \in Q : r < x} g( r) \). Since \( \sum_{r \in Q} g( r) \) is absolutely convergent, we know that \( f(x) \) is well-defined for every real number \( x \).

  1. Show that \( f \) is strictly monotone increasing. (Hint: you will need Proposition 5.4.14.)
  2. Show that for every rational number \( r \), \( f \) is discontinuous at \( r \). (Hint: since \( r \) is rational, \( r = q(n) \) for some natural number \( n \). Show that \( f(x) \geq f( r) + 2^{−n} \) for all \( x > r \).)
  3. Show that for every irrational number \( x \), \( f \) is continuous at \( x \). (Hint: first demonstrate that the functions \( f_n(x) := \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) \) are continuous at x, and that \( |f(x) - f_n(x)| \leq 2^{-n} \)).

证明1:

\( \forall x_1 < x_2 \in \mathbf{R} \),根据定理5.4.14,\( \exists q \in \mathbf{Q}, x_1 < q < x_2 \),可得\( f(x_2) = \sum_{r \in Q : r < x_2} g( r) = (\sum_{r \in Q : r < x_2, r \neq q} g( r)) + g(q) \),再由\( x_1 < q < x_2 \),可得\( \{ r \in Q : r < x_1 \} \subseteq \{ r \in Q : r < x_2, r \neq q \} \),因此\( f(x_1) = \sum_{r \in Q : r < x_1} g( r) \leq \sum_{r \in Q : r < x_2, r \neq q} g( r) \),进而\( f(x_1) < \sum_{r \in Q : r < x_2, r \neq q} g( r) + g(q) = f(x_2) \),即\( f \)严格单调递增。

证毕。

证明2:

\( \forall r \in \mathbf{Q} \),因为\( q \)双射,因此\( \exists n_0 \in \mathbf{N}, q(n_0) = r \),针对\( \forall x > r \),有\( f(x) = \sum_{p \in Q : p < x} g(p) = (\sum_{p \in Q : p < x, p \neq r} g(p)) + g( r) = (\sum_{p \in Q : p < x, p \neq r} g(p)) + g(q(n_0)) = (\sum_{p \in Q : p < x, p \neq r} g(p)) + 2^{-n_0} \),由于\( r < x \),因此\( \{ p \in Q : p < r \} \subseteq \{ p \in Q : p < x, p \neq r \} \),于是有\( \sum_{p \in Q : p < r} g(p) \leq \sum_{p \in Q : p < x, p \neq r} g(p) \),进而有\( (\sum_{p \in Q : p < r} g(p)) + 2^{-n_0} \leq (\sum_{p \in Q : p < x, p \neq r} g(p)) + 2^{-n_0} \),即\( f( r) + 2^{-n_0} \leq f(x) \) (1) 。令\( \epsilon_0 := \dfrac{2^{-n_0}}{2} \),针对\( \forall \delta > 0 \),令\( x_0 := r + \dfrac{\delta}{2} \),有\( |x_0 - r| < \delta \),因为\( x_0 > r \),因此由(1)可得, \( f(x_0) \geq f( r) + 2^{-n_0} \),进而\( |f(x_0) - f( r)| = f(x_0) - f( r) \geq 2^{-n_0} > \epsilon_0 \),根据定理9.4.7的3,可得\( f \)在\( r \)处不连续。

证毕。

证明3:

\( \forall n \in \mathbf{N} \),令函数\( f_n(x) := \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) \) (注:多了限制条件\( g( r) \geq 2^{-n} \),令\( r_n := q(n) \),则这相当于仅加位置在\( r_n \)前包括\( r_n \)本身的有理数(当然,还要求该有理数\( < x \)),这是因为\( 2^{-n} \)是严格单调递减函数),针对\( \forall x \in \mathbf{R} \setminus \mathbf{Q} \), \( \forall \epsilon > 0 \),令\( \delta := \min(\{ |x - q(i)| : 0 \leq i \leq n \}) \) (\( \delta \)即前\( n \)个有理数中,与\( x \)最近的那个有理数,它与\( x \)的距离,将间距\( \delta \)设为这个距离,则在这个间距内(开区间,不是闭区间)的有理数均不会是前\( n \)个有理数),由于\( x \notin \mathbf{Q} \),因此\( \delta > 0 \),针对\( \forall y \in \mathbf{R}, |y - x| < \delta \),有\( x - \delta < y < x + \delta \),接着分类讨论:

  1. 如果\( x - \delta < y \leq x \):此时\( \forall r \in \mathbf{Q}, y \leq r < x \) (即\( f(x) \)相比于\( f(y) \)额外加的那一部分),有\( r \notin \{ q(i) : 0 \leq i \leq n \} \) (1) (这是因为\( r \)与\( x \)的距离\( < \delta \),而\( \delta \)是前\( n \)个有理数中,与\( x \)距离最小的那个有理数,它与\( x \)间的距离),由\( q \)双射,可得\( \exists n_r \in \mathbf{N}, q(n_r) = r \),由(1)可得,\( n_r > n \),因此\( g( r) = 2^{-n_r} < 2^{-n} \),于是可得\( \{ r \in Q : y \leq r < x, g( r) \geq 2^{-n} \} = \emptyset \),进而\( \sum_{r \in Q : y \leq r < x, g( r) \geq 2^{-n}} g( r) = 0 \),这意味着\( f_n(x) = \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) = \sum_{r \in Q : r < y, g( r) \geq 2^{-n}} g( r) + \sum_{r \in Q : y \leq r < x, g( r) \geq 2^{-n}} g( r) = \sum_{r \in Q : r < y, g( r) \geq 2^{-n}} g( r) + 0 = \sum_{r \in Q : r < y, g( r) \geq 2^{-n}} g( r) = f_n(y) \),因此\( |f_n(y) - f_n(x)| = 0 < \epsilon \)。
  2. 如果\( x < y < x + \delta \):此时\( \forall r \in \mathbf{Q}, x \leq r < y \) 有\( r \notin \{ q(i) : 0 \leq i \leq n \} \) (2) 由\( q \)双射,可得\( \exists n_r \in \mathbf{N}, q(n_r) = r \),由(2)可得,\( n_r > n \),因此\( g( r) = 2^{-n_r} < 2^{-n} \),于是可得\( \{ r \in Q : x \leq r < y, g( r) \geq 2^{-n} \} = \emptyset \),进而\( \sum_{r \in Q : x \leq r < y, g( r) \geq 2^{-n}} g( r) = 0 \),这意味着\( f_n(y) = \sum_{r \in Q : r < y, g( r) \geq 2^{-n}} g( r) = \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) + \sum_{r \in Q : x \leq r < y, g( r) \geq 2^{-n}} g( r) = \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) + 0 = \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) = f_n(x) \),因此\( |f_n(y) - f_n(x)| = 0 < \epsilon \)。

综上,\( \forall x \in \mathbf{R} \setminus \mathbf{Q}, \exists \delta > 0 \), \( \forall y \in \mathbf{R}, |y - x| < \delta \),均有\( |f_n(y) - f_n(x)| < \epsilon \),由定理9.4.7的3,可得 \( f_n \)在\( x \)连续。至此可得,\( f_n \)在任意无理数点连续。

\( \forall n \in \mathbf{N} \),我们证明下\( \sum_{i = n + 1}^{\infty} g(q(i)) \leq 2^{-n} \):当\( n = 0 \),根据例子7.2.4,有\( \sum_{i = 1}^{\infty} 2^{-n} = 1 \),因此\( \sum_{i = n + 1}^{\infty} g(q(i)) = \sum_{i = 1}^{\infty} 2^{-n} = 1 \leq 1 = 2^{-n} \),基础情况成立。归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时,根据归纳假设,有\( \sum_{i = k + 1}^{\infty} g(q(i)) \leq 2^{-k} \),因此根据定理7.2.14的3,有\( \sum_{i = k + 2}^{\infty} g(q(i)) = (\sum_{i = k + 1}^{\infty} g(q(i))) - g(q(k + 1)) \leq 2^{-k} - 2^{-(k + 1)} = 2^{-k} - \dfrac{2^{-k}}{2} = \dfrac{2^{-k}}{2} = 2^{-(k + 1)} \),即\( n = k + 1 \)时也成立,归纳完毕。

\( \forall x \in \mathbf{R} \setminus \mathbf{Q} \),\( \forall n \in \mathbf{N} \),我们证明下\( |f(x) - f_n(x)| \leq 2^{-n} \):因为\( \{ r \in Q : r < x, g( r) < 2^{-n} \} = \{ q(i) : i > n, q(i) < x \} \),因此\( \sum_{r \in Q : r < x, g( r) < 2^{-n}} g( r) = \sum_{q(i) : i > n, q(i) < x} g(q(i)) \leq \sum_{i = n + 1}^{\infty} g(q(i)) \),前面证明了\( \sum_{i = n + 1}^{\infty} g(q(i)) \leq 2^{-n} \),因此\( \sum_{r \in Q : r < x, g( r) < 2^{-n}} g( r) \leq 2^{-n} \) (3) 。而\( f(x) = \sum_{r \in Q : r < x} g( r) = \sum_{r \in Q : r < x, g( r) \geq 2^{-n}} g( r) + \sum_{r \in Q : r < x, g( r) < 2^{-n}} g( r) = f_n(x) + \sum_{r \in Q : r < x, g( r) < 2^{-n}} g( r) \),结合(3),可得\( |f(x) - f_n(x)| = f(x) - f_n(x) = \sum_{r \in Q : r < x, g( r) < 2^{-n}} g( r) \leq 2^{-n} \)。

最后,我们证明\( f \)在任意无理数点连续,\( \forall x \in \mathbf{R} \setminus \mathbf{Q} \), \( \forall \epsilon > 0 \),有\( \exists n_0 \in \mathbf{N}, 0 < 2^{-n_0} \leq \dfrac{\epsilon}{3} \),根据前面的证明,有\( f_{n_0}(x) \)在\( x \)连续,根据定理9.4.7的4,这意味着 \( \exists \delta > 0, \forall y \in \mathbf{R}, |y - x| \leq \delta, |f_{n_0}(y) - f_{n_0}(x)| \leq \dfrac{\epsilon}{3} \),而根据前面的证明,有\( |f(y) - f_{n_0}(y)| \leq 2^{-{n_0}} \leq \dfrac{\epsilon}{3}, |f(x) - f_{n_0}(x)| \leq 2^{-{n_0}} \leq \dfrac{\epsilon}{3} \),根据定理4.3.7的3,有\( f(y) \)和\( f_{n_0}(x) \) \( \dfrac{2 \epsilon}{3} \)-接近,再次根据定理4.3.7的3以及\( f(x) \)和\( f_{n_0}(x) \) \( \dfrac{\epsilon}{3} \)-接近,可得\( f(y) \)和\( f(x) \) \( \epsilon \)-接近,即\( |f(y) - f(x)| \leq \epsilon \)。简而言之, \( \forall x \in \mathbf{R} \setminus \mathbf{Q} \), \( \forall \epsilon > 0, \exists \delta > 0, \forall y \in \mathbf{R}, |y - x| \leq \delta, |f(y) - f(x)| \leq \epsilon \),因此\( f \)在\( x \)连续。综上,\( f \)在任意无理数点连续。

证毕。

章节9.9

练习9.9.1

题目:

Prove Lemma 9.9.7.

Lemma 9.9.7的内容:

Let \( (a_n)_{n = 1}^{\infty} \) and \( (b_n)_{n = 1}^{\infty} \) be sequences of real numbers (not necessarily bounded or convergent). Then \( (a_n)_{n = 1}^{\infty} \) and \( (b_n)_{n = 1}^{\infty} \) are equivalent if and only if \( \lim_{n \to \infty} (a_n - b_n) = 0 \).

证明:

必要性:

如果\( (a_n)_{n = 1}^{\infty} \)和\( (b_n)_{n = 1}^{\infty} \)等价,根据定义9.9.5,可得\( \forall \epsilon > 0 \), \( (a_n)_{n = 1}^{\infty} \)和\( (b_n)_{n = 1}^{\infty} \) 均最终\( \epsilon \)-接近,即\( \exists N \geq 1 \) 使得\( (a_n)_{n = N}^{\infty} \)和\( (b_n)_{n = N}^{\infty} \) \( \epsilon \)-接近,于是有\( \forall n \geq N, |a_n - b_n| = |(a_n - b_n) - 0| \leq \epsilon \)。综上,有\( \lim_{n \to \infty} (a_n - b_n) = 0 \)。

充分性:

如果\( \lim_{n \to \infty} (a_n - b_n) = 0 \),则\( \forall \epsilon > 0, \exists N \geq 1, \forall n \geq N, |(a_n - b_n) - 0| = |a_n - b_n| \leq \epsilon \),可得\( (a_n)_{n = N}^{\infty} \)和\( (b_n)_{n = N}^{\infty} \) \( \epsilon \)-接近,进而有\( (a_n)_{n = 1}^{\infty} \)和\( (b_n)_{n = 1}^{\infty} \) 最终\( \epsilon \)-接近,根据定义9.9.5,这意味着 \( (a_n)_{n = 1}^{\infty} \)和\( (b_n)_{n = 1}^{\infty} \)等价。

证毕。

练习9.9.2

题目:

Prove Proposition 9.9.8. (Hint: you should avoid Lemma 9.9.7, and instead go back to the definition of equivalent sequences in Definition 9.9.5.)

Proposition 9.9.8的内容:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( f: X \to \mathbf{R} \) be a function. Then the following two statements are logically equivalent:

  1. \( f \) is uniformly continuous on \( X \).
  2. Whenever \( (x_n)_{n = 0}^{\infty} \) and \( (y_n)_{n = 0}^{\infty} \) are two equivalent sequences consisting of elements of \( X \), the sequences \( (f(x_n))_{n = 0}^{\infty} \) and \( (f(y_n))_{n = 0}^{\infty} \) are also equivalent.

证明:

\( 1 \Longrightarrow 2 \):

\( \forall \epsilon > 0 \),因为\( f \)在\( X \)上一致连续,因此\( \exists \delta > 0, \forall x, x_0 \in X, |x - x_0| \leq \delta \),有\( |f(x) - f(x_0)| \leq \epsilon \) (1) ,因为\( (x_n)_{n = 0}^{\infty}, (y_n)_{n = 0}^{\infty} \)等价,因此\( \exists N \geq 0, \forall n \geq N, |x_n - y_n| \leq \delta \),加上\( x_n, y_n \in X \),由(1)可得,\( |f(x_n) - f(y_n)| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists N \geq 0, \forall n \geq N, |f(x_n) - f(y_n)| \leq \epsilon \),即 \( (f(x_n))_{n = 0}^{\infty} \)和\( (f(y_n))_{n = 0}^{\infty} \)等价。

\( 2 \Longrightarrow 1 \):

假设\( f \)不在\( X \)上一致连续,即 \( \exists \epsilon_0 > 0, \forall \delta > 0 \),均\( \exists x, y \in X, |x - y| \leq \delta \),使得\( |f(x) - f(y)| > \epsilon_0 \) (1) 。 \( \forall n \in \mathbf{N} \),有\( \dfrac{1}{n + 1} > 0 \),根据(1)可得, \( \exists x_n, y_n \in X, |x_n - y_n| \leq \dfrac{1}{n + 1} \),使得\( |f(x_n) - f(y_n)| > \epsilon_0 \),于是有\( \forall n \in \mathbf{N}, \{ (x_n, y_n) : |x_n - y_n| \leq \dfrac{1}{n + 1}, |f(x_n) - f(y_n)| > \epsilon_0 \} \) 非空,使用选择公理,可以得到两个序列\( (x_n)_{n = 0}^{\infty}, (y_n)_{n = 0}^{\infty} \) 满足\( \forall n \in \mathbf{N}, |x_n - y_n| \leq \dfrac{1}{n + 1}, |f(x_n) - f(y_n)| > \epsilon_0 \) (2) ,易证\( (x_n)_{n = 0}^{\infty}, (y_n)_{n = 0}^{\infty} \)等价,但由(2)可得\( (f(x_n))_{n = 0}^{\infty}, (f(y_n))_{n = 0}^{\infty} \)不最终\( \epsilon_0 \)-接近(因为数列元素都\( > \epsilon_0 \)),进而\( (f(x_n))_{n = 0}^{\infty}, (f(y_n))_{n = 0}^{\infty} \)不等价,这和2矛盾,因此假设不成立,有\( f \)在\( X \)上一致连续。

证毕。

练习9.9.3

题目:

Prove Proposition 9.9.12. (Hint: use Definition 9.9.2 directly.)

Proposition 9.9.12的内容:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( f: X \to \mathbf{R} \) be a uniformly continuous function. Let \( (x_n)_{n = 0}^{\infty} \) be a Cauchy sequence consisting entirely of elements in \( X \). Then \( (f(x_n))_{n = 0}^{\infty} \) is also a Cauchy sequence.

证明:

\( \forall \epsilon > 0 \),因为\( f \)在\( X \)上一致连续,因此 \( \exists \delta > 0, \forall x, x_0 \in X, |x - x_0| \leq \delta, |f(x) - f(x_0)| \leq \epsilon \) (1) ,因为\( (x_n)_{n = 0}^{\infty} \)是柯西序列,因此 \( \exists N \geq 0, \forall j, k \geq N, |x_j - x_k| \leq \delta \),又\( x_j, x_k \in X \),加上(1),可得\( |f(x_j) - f(x_k)| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists N \geq 0, \forall j, k \geq N, |f(x_j) - f(x_k)| \leq \epsilon \),也就是\( (f(x_n))_{n = 0}^{\infty} \)也是柯西序列。

证毕。

练习9.9.4

题目:

Use Proposition 9.9.12 to prove Corollary 9.9.14. Use this corollary to give an alternate demonstration of the results in Example 9.9.10.

Corollary 9.9.14的内容:

Let \( X \) be a subset of \( \mathbf{R} \), let \( f: X \to \mathbf{R} \) be a uniformly continuous function, and let \( x_0 \) be an adherent point of \( X \). Then the limit \( \lim_{x \to x_0; x \in X} f(x) \) exists (in particular, it is a real number).

Example 9.9.10的内容:

Consider the function \( f: (0, 2) \to \mathbf{R} \) defined by \( f(x) := 1/x \) considered earlier. From Lemma 9.9.7 we see that the sequence \( (1/n)_{n = 1}^{\infty} \) and \( (1/2n)_{n = 1}^{\infty} \) are equivalent sequences in \( (0, 2) \). However, the sequences \( (f(1/n))_{n = 1}^{\infty} \) and \( (f(1/2n))_{n = 1}^{\infty} \) are not equivalent (why? Use Lemma 9.9.7 again). So by Proposition 9.9.8, \( f \) is not uniformly continuous. (These sequences start at 1 instead of 0, but the reader can easily see that this makes no difference to the above discussion.)

证明:

给定任意序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in X \)且\( \lim_{n \to \infty} a_n = x_0 \),由于\( (a_n)_{n = 0}^{\infty} \)收敛,因此它也是柯西序列,同时由于它的所有元素均\( \in X \),根据定理9.9.12,可得\( (f(a_n))_{n = 0}^{\infty} \)也是柯西序列,根据定理6.4.18, \( (f(a_n))_{n = 0}^{\infty} \)也收敛。

接下来我们想证明所有序列\( (f(a_n))_{n = 0}^{\infty} \)都收敛于同一个数,以使用定理9.3.9:针对\( \forall \)序列\( (b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty} \) 满足\( \forall n \in \mathbf{N}, b_n \in X, c_n \in X \)且 \( \lim_{n \to \infty} b_n = x_0, \lim_{n \to \infty} c_n = x_0 \),根据前面的证明,\( (f(b_n))_{n = 0}^{\infty}, (f(c_n))_{n = 0}^{\infty} \)均收敛,由于\( (b_n)_{n = 0}^{\infty}, (c_n)_{n = 0}^{\infty} \)均收敛于\( x_0 \),因此这两个序列等价,又它们的元素均\( \in X \),根据定理9.9.8,可得\( (f(b_n))_{n = 0}^{\infty}, (f(c_n))_{n = 0}^{\infty} \)也等价,特别的,它们也收敛于同一个数。

综上,根据定理9.3.9,可得\( \lim_{x \to x_0; x \in X} f(x) \)存在。

证毕。

给出另一种展示Example 9.9.10中的\( f \)不在\( (0, 2) \)上一致连续的方法:

\( 0 \)是\( (0, 2) \)的附着点,但\( x \to 0 \)时, \( f(x) \to +\infty \),故明显\( \lim_{x \to 0; x \in (0, 2)} f(x) \)不存在,进而根据推论9.9.14,可得\( f \)不在\( (0, 2) \)上一致连续。

练习9.9.5

题目:

Prove Proposition 9.9.15. (Hint: mimic the proof of Lemma 9.6.3. At some point you will need either Proposition 9.9.12 or Corollary 9.9.14.)

Proposition 9.9.15的内容:

Let \( X \) be a subset of \( \mathbf{R} \), and let \( f: X \to \mathbf{R} \) be a uniformly continuous function. Suppose that \( E \) is a bounded subset of \( X \). Then \( f(E) \) is also bounded.

证明:

假设\( f(E) \)无界,则\( \forall n \in \mathbf{N}, \exists x \in E, |f(x)| \geq n \),令\( X_n := \{ x \in E, |f(x)| \geq n \} \),有\( X_n \)非空,使用选择公理,可以得到序列\( (x_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, x_n \in X_n \),于是有\( x_n \in E \)且\( |f(x_n)| \geq n \) (1) ,令\( X := \{ x_n : n \in \mathbf{N} \} \),有\( X \subseteq E \),又\( E \)有界,因此\( X \)有界,这意味着\( (x_n)_{n = 0}^{\infty} \)有界,根据定理6.6.8,\( (x_n)_{n = 0}^{\infty} \)有收敛的子序列\( (x_{n_j})_{j = 0}^{\infty} \),特别的,有\( (x_{n_j})_{j = 0}^{\infty} \)是柯西序列,因为\( (x_{n_j})_{j = 0}^{\infty} \)的元素均\( \in E \),进而均\( \in X \),根据定理9.9.12,有\( (f(x_{n_j}))_{j = 0}^{\infty} \)也是柯西序列,进而有\( (f(x_{n_j}))_{j = 0}^{\infty} \) 有界 (2) ,根据(1)可得,\( \forall j \in \mathbf{N}, |f(x_{n_j})| \geq n_j \),这意味着\( (f(x_{n_j}))_{j = 0}^{\infty} \) 无界,这和(2)矛盾,因此假设不成立,有\( f(E) \)有界。

证毕。

练习9.9.6

题目:

Let \( X, Y, Z \) be subsets of \( \mathbf{R} \). Let \( f: X \to Y \) be a function which is uniformly continuous on \( X \), and let \( g: Y \to Z \) be a function which is uniformly continuous on \( Y \). Show that the function \( g \circ f : X \to Z \) is uniformly continuous on \( X \).

证明:

\( \forall \epsilon > 0 \),因为\( g \)在\( Y \)上一致连续,因此\( \exists \delta_0 > 0, \forall y, y_0 \in Y, |y - y_0| \leq \delta_0, |g(y) - g(y_0)| \leq \epsilon \) (1) ,因为\( f \)在\( X \)上一致连续,因此\( \exists \delta > 0, \forall x, x_0 \in X, |x - x_0| \leq \delta, |f(x) - f(x_0)| \leq \delta_0 \),进而根据(1)有,\( |g(f(x)) - g(f(x_0))| = |(g \circ f)(x) - (g \circ f)(x_0)| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists \delta > 0, \forall x, x_0 \in X, |x - x_0| \leq \delta, |(g \circ f)(x) - (g \circ f)(x_0)| \leq \epsilon \),即\( g \circ f \)在\( X \)上一致连续。

证毕。

章节9.10

练习9.10.1

题目:

Let \( (a_n)_{n = 0}^{\infty} \) be a sequence of real numbers, then \( a_n \) can also be thought of as a function from \( \mathbf{N} \) to \( \mathbf{R} \), which takes each natural number \( n \) to a real number \( a_n \). Show that \( \lim_{n \to +\infty; n \in \mathbf{N}} a_n = \lim_{n \to \infty} a_n \) where the left-hand limit is defined by Definition 9.10.3 and the right-hand limit is defined by Definition 6.1.8. More precisely, show that if one of the above two limits exists then so does the other, and then they both have the same value. Thus the two notions of limit here are compatible.

证明:

必要性:

如果\( \lim_{n \to +\infty; n \in \mathbf{N}} a_n = L \),我们要证明\( \lim_{n \to \infty} a_n = L \):因为\( \lim_{n \to +\infty; n \in \mathbf{N}} a_n = L \),因此\( \forall \epsilon > 0, \exists M \in \mathbf{R}, \forall n \in \mathbf{N} \cap (M, +\infty) \),有\( |a_n - L| \leq \epsilon \),这意味着\( \forall n \in \mathbf{N}, n > M \),有\( |a_n - L| \leq \epsilon \)。又\( \exists N \in \mathbf{N}, N > M \),可得\( \forall n \in \mathbf{N}, n \geq N \),有\( n > M \),进而有\( |a_n - L| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists N \in \mathbf{N}, \forall n \geq N, |a_n - L| \leq \epsilon \),也就是\( \lim_{n \to \infty} a_n = L \)。

充分性:

如果\( \lim_{n \to \infty} a_n = L \),我们要证明 \( \lim_{n \to +\infty; n \in \mathbf{N}} a_n = L \):因为\( \lim_{n \to \infty} a_n = L \),因此\( \forall \epsilon > 0, \exists N \in \mathbf{N}, \forall n \geq N, |a_n - L| \leq \epsilon \),取\( M := N \),则\( \forall n \in \mathbf{N} \cap (M, +\infty) \),有\( n > M = N \),进而有\( |a_n - L| \leq \epsilon \)。简而言之,\( \forall \epsilon > 0, \exists M \in \mathbf{R}, \forall n \in \mathbf{N} \cap (M, +\infty) \),有\( |a_n - L| \leq \epsilon \),也就是 \( \lim_{n \to +\infty; n \in \mathbf{N}} a_n = L \)。

证毕。

参考文章

  1. Functions from topological spaces to complete lattices
  2. Limit Superior and Limit Inferior of Functions
  3. What is the limit infimum and limit supremum of a function?中George Mathew的回答