目录

陶哲轩Analysis I习题的参考解答及思考(第11章)

第11章

版本

Analysis I(第3版)。

章节11.1

练习11.1.1

题目:

Prove Lemma 11.1.4. (Hint: in order to show that 1 implies 2 in the case when \( X \) is non-empty, consider the supremum and infimum of \( X \).)

Lemma 11.1.4的内容:

Let \( X \) be a subset of the real line. Then the following two statements are logically equivalent:

  1. \( X \) is bounded and connected.
  2. \( X \) is a bounded interval.

证明:

证明\( 1 \Longrightarrow 2 \):

如果\( X \)是有界且连通的:

如果\( X = \emptyset \)(\( \emptyset \)是有界且连通的),则\( X = (0, 0) \)为有界区间(或者也可以\( X = [1, 0] \)等等,区间的定义见定义9.1.1)。

如果\( X \neq \emptyset \),则由\( X \)有界,可得\( \sup(X), \inf(X) \)均为实数,下面分情况讨论:

  1. 如果\( \sup(X), \inf(X) \)均\( \notin X \) (1) ,我们证明\( X = (\inf(X), \sup(X)) \):\( \forall z \in (\inf(X), \sup(X)) \),有\( \inf(X) < z < \sup(X) \),进而\( \exists x \in X, y \in X \)满足 \( x < z < y \)(\( z \)小于\( X \)的上确界,于是\( X \)中存在元素大于\( z \)(否则上确界应该更小,矛盾),下确界同理),再由\( X \)是连通的,可得\( [x, y] \subseteq X \),而\( z \in [x, y] \),于是有\( z \in X \),综上,有\( (\inf(X), \sup(X)) \subseteq X \)。 \( \forall z \in X \),有\( \inf(X) \leq z \leq \sup(X) \),又(1),于是有\( \inf(X) < z < \sup(X) \),可得\( z \in (\inf(X), \sup(X)) \),综上,有\( X \subseteq (\inf(X), \sup(X)) \),结合前面的\( (\inf(X), \sup(X)) \subseteq X \),可得\( X = (\inf(X), \sup(X)) \),于是有\( X \)为有界区间。
  2. 如果\( \sup(X), \inf(X) \)均\( \in X \),我们证明\( X = [\inf(X), \sup(X)] \):\( \forall z \in [\inf(X), \sup(X)] \),有\( \inf(X) \leq z \leq \sup(X) \),进而\( \exists x \in X, y \in X \)满足 \( x \leq z \leq y \),再由\( X \)是连通的,可得\( [x, y] \subseteq X \),而\( z \in [x, y] \),于是有\( z \in X \),综上,有\( [\inf(X), \sup(X)] \subseteq X \)。 \( \forall z \in X \),有\( \inf(X) \leq z \leq \sup(X) \),可得\( z \in [\inf(X), \sup(X)] \),综上,有\( X \subseteq [\inf(X), \sup(X)] \),结合前面的\( [\inf(X), \sup(X)] \subseteq X \),可得\( X = [\inf(X), \sup(X)] \),于是有\( X \)为有界区间。
  3. 如果\( \inf(X) \notin X, \sup(X) \in X \),我们证明\( X = (\inf(X), \sup(X)] \):\( \forall z \in (\inf(X), \sup(X)] \),有\( \inf(X) < z \leq \sup(X) \),进而\( \exists x \in X, y \in X \)满足 \( x < z \leq y \),再由\( X \)是连通的,可得\( [x, y] \subseteq X \),而\( z \in [x, y] \),于是有\( z \in X \),综上,有\( (\inf(X), \sup(X)] \subseteq X \)。 \( \forall z \in X \),有\( \inf(X) \leq z \leq \sup(X) \),又\( \inf(X) \notin X \),于是有\( \inf(X) < z \leq \sup(X) \),可得\( z \in (\inf(X), \sup(X)] \),综上,有\( X \subseteq (\inf(X), \sup(X)] \),结合前面的\( (\inf(X), \sup(X)] \subseteq X \),可得\( X = (\inf(X), \sup(X)] \),于是有\( X \)为有界区间。
  4. 如果\( \inf(X) \in X, \sup(X) \notin X \),这种情况跟情况3类似,易得\( X \)为有界区间\( [\inf(X), \sup(X)) \)。

综上,有\( X \)为有界区间。

证明\( 2 \Longrightarrow 1 \):

如果\( X \)为有界区间,则\( X \)是有界的,还差证明\( X \)连通,由\( X \)为有界区间,可得\( \exists a, b \in \mathbf{R} \),使得\( X = (a, b) \)或\( X = [a, b] \)或\( X = (a, b] \)或\( X = [a, b) \),分情况讨论,易证所有情况下,\( X \)均是连通的。

证毕。

练习11.1.2

题目:

Prove Corollary 11.1.6. (Hint: use Lemma 11.1.4, and explain why the intersection of two bounded sets is automatically bounded, and why the intersection of two connected sets is automatically connected.)

Corollary 11.1.6的内容:

If \( I \) and \( J \) are bounded intervals, then the intersection \( I \cap J \) is also a bounded interval.

证明:

由\( I, J \)均是有界区间以及引理11.1.4,可得\( I, J \)均有界且连通。

由\( I, J \)均有界,可得\( \exists M_I \in \mathbf{R}, \forall x \in I, |x| \leq M_I \) 以及\( \exists M_J \in \mathbf{R}, \forall x \in J, |x| \leq M_J \),取\( M := \max(M_I, M_J) \),可得\( \forall x \in I \cap J, |x| \leq M \),因此\( I \cap J \)也有界。

\( \forall x < y \in I \cap J \),有\( x, y \in I \)且\( x, y \in J \),再由\( I, J \)连通,可得\( [x, y] \subseteq I, [x, y] \subseteq J \),进而可得\( [x, y] \subseteq I \cap J \),综上,有\( I \cap J \)也连通。

由\( I \cap J \)有界且连通以及引理11.1.4,可得\( I \cap J \)为有界区间。

证毕。

练习11.1.3

题目:

Let \( I \) be a bounded interval of the form \( I = (a, b) \) or \( I = [a, b) \) for some real numbers \( a < b \). Let \( I_1, \dots, I_n \) be a partition of \( I \). Prove that one of the intervals \( I_j \) in this partition is of the form \( I_j = (c, b) \) or \( I_j = [c, b) \) for some \( a \leq c \leq b \). (Hint: prove by contradiction. First show that if \( I_j \) is not of the form \( (c, b) \) or \( [c, b) \) for any \( a \leq c \leq b \), then \( \sup(I_j) \) is strictly less than \( b \).)

证明:

假设\( \forall 1 \leq j \leq n \),均有\( \forall a \leq c \leq b \), \( I_j \neq (c, b), I_j \neq [c, b) \) (1) ,我们证明\( \sup(I_j) < b \):

  1. 如果\( I_j = \emptyset \),则\( \sup(I_j) = -\infty < b \),
  2. 如果\( I_j \)为单例集,即\( \exists x \in I, I_j = \{ x \} \),此时由于\( I_j \subseteq I \)以及\( I = (a, b) \)或\( I = [a, b) \),可得\( x < b \),于是\( \sup(I_j) = x < b \)。
  3. 如果\( I_j \)不为空集也不为单例集,此时\( \exists l < r \in \mathbf{R} \),使得\( I_j = \)四种形式\( (l, r), [l, r], (l, r], [l, r) \)中的其中一种(\( I_j \)是分区中的一个集合,因此是一个区间),此时由\( I_j \subseteq I \)以及\( I = (a, b) \)或\( I = [a, b) \),易得\( a \leq r \leq b \)(重点注意\( l < r \),然后考虑下,如果\( r < a \)会怎么样,如果\( r > b \)呢),再由(1),可得\( r \neq b \),于是有\( a \leq r < b \),又不管\( I_j \)是四种形式中的哪一种,均有\( \sup(I_j) = r \),于是有\( \sup(I_j) < b \)。

综上,有\( \sup(I_j) < b \)。

令\( M := \max(\{ \sup(I_j) : 1 \leq j \leq n \}) \) (分区是有限集,因此可以取最大值),可得\( M < b \) 且\( \forall x \in I, x \leq M \)(也就是\( M \)为\( I \)的上界),于是有\( \sup(I) \leq M < b \),但\( \sup(I) = b \),矛盾,因此假设不成立, \( \exists 1 \leq j_0 \leq n \)满足\( \exists a \leq c_0 \leq b \),使得\( I_{j_0} = (c_0, b) \)或\( I_{j_0} = [c_0, b) \)。

证毕。

练习11.1.4

题目:

Prove Lemma 11.1.18.

Lemma 11.1.18的内容:

Let \( I \) be a bounded interval, and let \( \mathbf{P} \) and \( \mathbf{P'} \) be two partitions of \( I \). Then \( \mathbf{P} \# \mathbf{P'} \) is also a partition of \( I \), and is both finer than \( \mathbf{P} \) and finer than \( \mathbf{P'} \).

证明:

证明\( \mathbf{P} \# \mathbf{P'} \)也是\( I \)的分区:

先证明\( \bigcup_{A \in \mathbf{P} \# \mathbf{P'}} A = I \):\( \forall x \in \bigcup_{A \in \mathbf{P} \# \mathbf{P'}} A \),有\( \exists A_x \in \mathbf{P} \# \mathbf{P'}, x \in A_x \),而根据\( \mathbf{P} \# \mathbf{P'} \)的定义,可得 \( \exists K_x \in \mathbf{P}, J_x \in \mathbf{P'} \),使得\( K_x \cap J_x = A_x \),又\( K_x \subseteq I \),因此\( A_x \subseteq I \),于是可得\( x \in I \)。反之,\( \forall x \in I \),因为\( \mathbf{P}, \mathbf{P'} \)是\( I \)的分区,因此\( \exists K_x \in \mathbf{P}, x \in K_x, \exists J_x \in \mathbf{P'}, x \in J_x \),于是有\( x \in K_x \cap J_x \),进而根据\( \mathbf{P} \# \mathbf{P'} \)的定义,可得\( x \in \bigcup_{A \in \mathbf{P} \# \mathbf{P'}} A \)。综上,有\( \bigcup_{A \in \mathbf{P} \# \mathbf{P'}} A = I \)。

接着证明\( \mathbf{P} \# \mathbf{P'} \)中的集合互不相交:\( \forall A, B \in \mathbf{P} \# \mathbf{P'} \),根据\( \mathbf{P} \# \mathbf{P'} \)的定义,可得\( \exists K_A, K_B \in \mathbf{P}, J_A, J_B \in \mathbf{P'} \),使得\( A = K_A \cap J_A, B = K_B \cap J_B \),假设\( A \cap B \neq \emptyset \),则\( \exists x_0 \in A \cap B \),特别的,有\( x_0 \in K_A \)且\( x_0 \in K_B \),于是有\( K_A \cap K_B \neq \emptyset \),但这和\( \mathbf{P} \)是\( I \)的分区矛盾,因此假设不成立,有\( A \cap B = \emptyset \)。

综上,\( \forall x \in I \),\( \exists \)唯一的\( A \in \mathbf{P} \# \mathbf{P'} \),使得\( x \in A \),即\( \mathbf{P} \# \mathbf{P'} \)也是\( I \)的分区。

证明\( \mathbf{P} \# \mathbf{P'} \)比\( \mathbf{P}, \mathbf{P'} \)都精细:

\( \forall A \in \mathbf{P} \# \mathbf{P'}, \exists K_A \in \mathbf{P}, J_A \in \mathbf{P'} \),使得\( A = K_A \cap J_A \),于是有\( A \subseteq K_A, A \subseteq K_B \)。综上,有\( \mathbf{P} \# \mathbf{P'} \)比\( \mathbf{P}, \mathbf{P'} \)都精细。

证毕。

章节11.2

练习11.2.1

题目:

Prove Lemma 11.2.7.

Lemma 11.2.7的内容:

Let \( I \) be a bounded interval, let \( \mathbf{P} \) be a partition of \( I \), and let \( f: I \to \mathbf{R} \) be a function which is piecewise constant with respect to \( \mathbf{P} \). Let \( \mathbf{P'} \) be a partition of \( I \) which is finer than \( \mathbf{P} \). Then \( f \) is also piecewise constant with respect to \( \mathbf{P'} \).

证明:

\( \forall J \in \mathbf{P'} \),因为\( \mathbf{P'} \)比\( \mathbf{P} \)精细,因此\( \exists K \in \mathbf{P}, J \subseteq K \),又\( f \)针对\( \mathbf{P} \)为分段常数函数,因此\( \exists c \in \mathbf{R}, \forall x \in K, f(x) = c \),而\( J \subseteq K \),因此\( \forall x \in J \),有\( x \in K \),进而也有\( f(x) = c \)。综上,\( f \)针对\( \mathbf{P'} \)也为分段常数函数。

证毕。

练习11.2.2

题目:

Prove Lemma 11.2.8. (Hint: use Lemmas 11.1.18 and 11.2.7 to make \( f \) and \( g \) piecewise constant with respect to the same partition of \( I \).)

Lemma 11.2.8的内容:

Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) be piecewise constant functions on \( I \). Then the functions \( f + g, f − g, \max(f, g) \) and \( fg \) are also piecewise constant functions on \( I \). Here of course \( \max(f, g): I \to \mathbf{R} \) is the function \( \max(f, g)(x) := \max(f (x), g(x)) \). If \( g \) does not vanish anywhere on \( I \) (i.e., \( g(x) \neq 0 \) for all \( x \in I \)) then \( f/g \) is also a piecewise constant function on \( I \).

注:

这里作者没让证明\( \min(f, g) \)的情况,实际上\( \min(f, g) \)也为 \( I \)上的分段常数函数,这里一起证了。

证明:

因为\( f \)和\( g \)为\( I \)上的分段常数函数,因此 \( \exists I \)的分区\( \mathbf{P}, \mathbf{P'} \)使得\( f \)针对\( \mathbf{P} \)为分段常数函数, \( g \)针对\( \mathbf{P'} \)为分段常数函数。根据引理11.1.18, \( \mathbf{P} \# \mathbf{P'} \)也是\( I \)的分区且\( \mathbf{P} \)比\( \mathbf{P}, \mathbf{P'} \)都精细,此时再根据引理11.2.7,可得\( f, g \)均针对\( \mathbf{P} \# \mathbf{P'} \)为分段常数函数,于是\( \forall J \in \mathbf{P} \# \mathbf{P'}, \exists c_1, c_2 \in \mathbf{R}, \forall x \in J, f(x) = c_1, g(x) = c_2 \),进而\( (f + g)(x) = f(x) + g(x) = c_1 + c_2, (f - g)(x) = f(x) - g(x) = c_1 - c_2, \max(f, g)(x) = \max(f(x), g(x)) = \max(c_1, c_2), \min(f, g)(x) = \min(f(x), g(x)) = \min(c_1, c_2), (fg)(x) = f(x)g(x) = c_1c_2 \),这里\( c_1 + c_2, c_1 - c_2, \max(c_1, c_2), \min(c_1, c_2), c_1c_2 \)均为常数,若\( g \)进一步满足\( \forall x \in I, g(x) \neq 0 \),则\( c_2 \neq 0 \),此时有\( (f/g)(x) = f(x)/g(x) = c_1/c_2 \)也为常数。综上,\( f + g, f - g, \max(f, g), \min(f, g), fg \)均针对\( \mathbf{P} \# \mathbf{P'} \)为分段常数函数且如果\( \forall x \in I, g(x) \neq 0 \),则\( f/g \)也针对\( \mathbf{P} \# \mathbf{P'} \)为分段常数函数,于是,\( f + g, f - g, \max(f, g), \min(f, g), fg \)均为\( I \)上的分段常数函数且如果\( \forall x \in I, g(x) \neq 0 \),则\( f/g \)也为\( I \)上的分段常数函数。

证毕。

练习11.2.3

题目:

Prove Proposition 11.2.13. (Hint: first use Theorem 11.1.13 to show that both integrals are equal to \( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \).)

Proposition 11.2.13的内容:

(Piecewise constant integral is independent of partition). Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) be a function. Suppose that \( \mathbf{P} \) and \( \mathbf{P'} \) are partitions of \( I \) such that \( f \) is piecewise constant both with respect to \( \mathbf{P} \) and with respect to \( \mathbf{P'} \). Then \( p.c. \int_{[\mathbf{P}]} f = p.c. \int_{[\mathbf{P'}]} f \).

注:

给定一个区间\( J \)(或者其他区间符号),如果\( f \)在\( J \)上为常数函数,则我们约定\( c_J \)为\( f \)在\( J \)上的常数值,不再显式引入符号\( c_J \)。

证明:

因为\( \mathbf{P}, \mathbf{P'} \)均是\( I \)的分区,根据引理11.1.18,可得\( \mathbf{P} \# \mathbf{P'} \)也是\( I \)的分区且比\( \mathbf{P}, \mathbf{P'} \)都精细。

\( \forall J \in \mathbf{P} \),令\( \mathbf{P_J} := \{ J \cap K : K \in \mathbf{P'} \} \),我们证明\( \mathbf{P_J} \)为\( J \)的分区:首先证明\( \bigcup_{L \in \mathbf{P_J}} L = J \),分情况讨论:

  1. 如果\( J = \emptyset \),则\( \bigcup_{L \in \mathbf{P_J}} L = \emptyset = J \)。
  2. 如果\( J \neq \emptyset \),则根据\( \mathbf{P_J} \)的定义,明显有\( \bigcup_{L \in \mathbf{P_J}} L \subseteq J \),再证明下\( J \subseteq \bigcup_{L \in \mathbf{P_J}} L \)就行,因为\( \mathbf{P} \# \mathbf{P'} \)比\( \mathbf{P} \)精细,因此\( \exists J_L \in \mathbf{P}, K_L \in \mathbf{P'}, J \subseteq J_L \cap K_L \),又\( \mathbf{P} \)中的区间互不相交,因此\( J_L = J \)(注意,\( J \neq \emptyset \)),于是有\( J \subseteq J \cap K_L \),加上\( J \cap K_L \subseteq \bigcup_{L \in \mathbf{P_J}} L \),可得\( J \subseteq \bigcup_{L \in \mathbf{P_J}} L \),综上,有\( \bigcup_{L \in \mathbf{P_J}} L = J \)。

接着证明\( \mathbf{P_J} \)中的区间互不相交,假设\( \exists L_1, L_2 \in \mathbf{P_J}, L_1 \cap L_2 \neq \emptyset \),则由于\( \mathbf{P_J} \subseteq \mathbf{P} \# \mathbf{P'} \),可得\( L_1, L_2 \in \mathbf{P} \# \mathbf{P'} \),而\( L_1 \cap L_2 \neq \emptyset \),这与\( \mathbf{P} \# \mathbf{P'} \)是\( I \)的分区矛盾, 因此假设不成立,有\( \mathbf{P_J} \)中的区间互不相交。

综上,有\( \mathbf{P_J} \)为\( J \)的分区。

根据引理7.1.13,可得\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f = \sum_{L \in \mathbf{P} \# \mathbf{P'}} c_L|L| = \sum_{J \in \mathbf{P}} (\sum_{K \in \mathbf{P'}} c_{J \cap K}|J \cap K|) \) (1)

\( \forall J \in \mathbf{P} \),因为\( f \)针对\( \mathbf{P} \)为分段常数函数,因此\( f \)在\( J \)上为常数函数(注意,没有“分段”两个字) (2) ,针对\( \forall K \in \mathbf{P'} \),由\( J \cap K \subseteq J \)以及(2),可得\( c_{J \cap K} = c_J \),于是有\( \sum_{K \in \mathbf{P'}} c_{J \cap K}|J \cap K| = \sum_{K \in \mathbf{P'}} c_J|J \cap K| \),而\( \sum_{K \in \mathbf{P'}} c_J|J \cap K| = \sum_{L \in \mathbf{P_J}} c_J|L| \),因此可得\( \sum_{K \in \mathbf{P'}} c_{J \cap K}|J \cap K| = \sum_{L \in \mathbf{P_J}} c_J|L| = c_J \sum_{L \in \mathbf{P_J}} |L| \),再根据定理11.1.13以及\( \mathbf{P_J} \)为\( J \)的分区,可得\( \sum_{L \in \mathbf{P_J}} |L| = |J| \),于是有\( \sum_{K \in \mathbf{P'}} c_{J \cap K}|J \cap K| = c_J \sum_{L \in \mathbf{P_J}} |L| = c_J|J| \) (3)

根据(1)和(3),\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f = \sum_{J \in \mathbf{P}} (\sum_{K \in \mathbf{P'}} c_{J \cap K}|J \cap K|) = \sum_{J \in \mathbf{P}} c_J|J| = p.c. \int_{[\mathbf{P}]} f \)。

同理易证\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f = p.c. \int_{[\mathbf{P'}]} f \),于是可得\( p.c. \int_{[\mathbf{P}]} f = p.c. \int_{[\mathbf{P'}]} f \)。

证毕。

练习11.2.4

题目:

Prove Theorem 11.2.16. (Hint: you can use earlier parts of the theorem to prove some of the later parts of the theorem. See also the hint to Exercise 11.2.2.)

Theorem 11.2.16的内容:

(Laws of integration). Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) be piecewise constant functions on \( I \).

  1. We have \( p.c. \int_{I} (f + g) = p.c. \int_{I} f + p.c. \int_{I} g \).
  2. For any real number \( c \), we have \( p.c. \int_{I} (cf) = c (p.c. \int_{I} f) \).
  3. We have \( p.c. \int_{I} (f - g) = p.c. \int_{I} f - p.c. \int_{I} g \).
  4. If \( f(x) \geq 0 \) for all \( x \in I \), then \( p.c. \int_{I} f \geq 0 \).
  5. If \( f(x) \geq g(x) \) for all \( x \in I \), then \( p.c. \int_{I} f \geq p.c. \int_{I} g \).
  6. If \( f \) is the constant function \( f(x) = c \) for all \( x \in I \), then \( p.c. \int_{I} f = c|I| \).
  7. Let \( J \) be a bounded interval containing \( I \) (i.e. \( I \subseteq J \)), and let \( F: J \to \mathbf{R} \) be the function \( F(x) := \begin{cases} f(x) & \text{ if } x \in I \\ 0 & \text{ if } x \notin I \end{cases} \), Then \( F \) is piecewise constant on \( J \), and \( p.c. \int_{J} F = p.c. \int_{I} f \).
  8. Suppose that \( \{ J, K \} \) is a partition of \( I \) into two intervals \( J \) and \( K \). Then the functions \( f|_J: J \to \mathbf{R} \) and \( f|_K: K \to \mathbf{R} \) are piecewise constant on \( J \) and \( K \) respectively, and we have \( p.c. \int_{I} f = p.c. \int_{J} f|_J + p.c. \int_{K} f|_K \).

注:

证明过程中会隐式用到定理11.2.13,为了写得比较流畅点,没有明确指出来,读者自己注意下。

证明:

因为\( f, g \)均为\( I \)上的分段常数函数,因此\( \exists I \)的分区\( \mathbf{P}, \mathbf{P'} \)使得 \( f \)针对\( \mathbf{P} \)为分段常数函数, \( g \)针对\( \mathbf{P'} \)为分段常数函数,根据引理11.1.18和引理11.2.7,可得\( f, g \)均针对\( \mathbf{P} \# \mathbf{P'} \)为分段常数函数, \( \forall K \in \mathbf{P} \# \mathbf{P'} \),令\( c_{K_f} \)为\( f \)在\( K \)上的常数值,令\( c_{K_g} \)为\( g \)在\( K \)上的常数值。

证明1:

\( p.c. \int_{I} (f + g) = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} (f + g) = \sum_{K \in \mathbf{P} \# \mathbf{P'}} (c_{K_f} + c_{K_g})|K| = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f}|K|) + (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_g}|K|) = (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f) + (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} g) = (p.c. \int_{I} f) + (p.c. \int_{I} g) \)。

证明2:

\( \forall c \in \mathbf{R}, p.c. \int_{I} (cf) = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} (cf) = \sum_{K \in \mathbf{P} \# \mathbf{P'}} cc_{K_f}|K| = c (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f}|K|) = c (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f) = c (p.c. \int_{I} f) \)。

证明3:

根据2,有\( p.c. \int_{I} (-g) = -(p.c. \int_{I} g) \),再根据1,可得\( p.c. \int_{I} (f - g) = p.c. \int_{I} (f + (-g)) = (p.c. \int_{I} f) + (p.c. \int_{I} (-g)) = (p.c. \int_{I} f) - (p.c. \int_{I} g) \)。

证明4:

如果\( \forall x \in I, f(x) \geq 0 \),则\( \forall K \in \mathbf{P} \# \mathbf{P'}, c_{K_f} \geq 0 \),此时\( p.c. \int_{I} f = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f}|K| \geq 0 \)。

证明5:

如果\( \forall x \in I, f(x) \geq g(x) \),则\( \forall x \in I, (f - g)(x) = f(x) - g(x) \geq 0 \),此时根据4以及3,可得\( p.c. \int_{I} (f - g) = (p.c. \int_{I} f) - (p.c. \int_{I} g) \geq 0 \),进而可得\( p.c. \int_{I} f \geq p.c. \int_{I} g \)。

证明6:

由\( \mathbf{P} \# \mathbf{P'} \)是\( I \)的分区以及定理11.1.13,可得 \( \sum_{K \in \mathbf{P} \# \mathbf{P'}} |K| = |I| \)。

如果\( \exists c \in \mathbf{R}, \forall x \in I, f(x) = c \),则\( \forall K \in \mathbf{P} \# \mathbf{P'}, c_{K_f} = c \),此时\( p.c. \int_{I} f = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f}|K| = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c|K| = c (\sum_{K \in \mathbf{P} \# \mathbf{P'}} |K|) = c|I| \)。

证明7:

如果\( I = \emptyset \),则\( \forall x \in J, F(x) = 0 \),此时易得\( p.c. \int_{J} F = p.c. \int_{I} f = 0 \)。

下面考虑\( I \neq \emptyset \)的情况:

因为\( I \)为有界区间以及\( I \neq \emptyset \),因此\( \exists a_I \leq b_I \in \mathbf{R} \) 使得\( I \)为\( [a_I, b_I], (a_I, b_I), [a_I, b_I), (a_I, b_I] \)中的其中一种。因为\( I \subseteq J \),因此\( J \neq \emptyset \),又因为\( J \)为有界区间,因此\( \exists a_J \leq b_J \in \mathbf{R} \) 使得\( J \)为\( [a_J, b_J], (a_J, b_J), [a_J, b_J), (a_J, b_J] \)中的其中一种。因为\( I \subseteq J \),因此\( a_J \leq a_I, b_I \leq b_J \) (考虑下,如果\( a_J > a_I \)或\( b_I > b_J \),则不管两个区间是哪种形式,均会违反\( I \subseteq J \)),可得\( J - I = J_l \cup J_r \),其中\( J_l, J_r \)均为有界区间,\( J_l \cap J_r = \emptyset \),并且,\( J_l \)为\( [a_J, a_I], (a_J, a_I), [a_J, a_I), (a_J, a_I] \)中的其中一种, \( J_r \)为\( [b_I, b_J], (b_I, b_J), [b_I, b_J), (b_I, b_J] \)中的其中一种,令\( \mathbf{P_J} := (\mathbf{P} \# \mathbf{P'}) \cup \{ J_l, J_r \} \),易得\( \mathbf{P_J} \)为\( J \)的分区且\( F \)针对\( \mathbf{P_J} \)为分段常数函数。 \( \forall K \in \mathbf{P_J} \),令\( c_{K_F} \)为\( F \)在\( K \)上的常数值。 \( \forall x \in J_l \cup J_r = J - I \),有\( x \notin I \),因此\( F(x) = 0 \),进而可得\( F \)在\( J_l, J_r \)上的常数值\( c_{J_l}, c_{J_r} \)均\( = 0 \)。 \( \forall x \in \bigcup_{K \in \mathbf{P} \# \mathbf{P'}} K \),有\( x \in I \),因此\( F(x) = f(x) \),进而可得\( \forall K \in \mathbf{P} \# \mathbf{P'} \), \( F \)在\( K \)上的常数值\( c_{K_F} = \)\( f \)在\( K \)上的常数值\( c_{K_f} \),此时可得\( p.c. \int_{J} F = p.c. \int_{[\mathbf{P_J}]} F = \sum_{K \in \mathbf{P_J}} c_{K_F}|K| = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F}|K|) + (c_{J_l}|J_l| + c_{J_r}|J_r|) = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F}|K|) + (0|J_l|) + (0|J_r|) = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F}|K| = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f}|K| = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f = p.c. \int_{I} f \)。

证明8:

注:这里为了方便,我们shadow之前定义的\( \mathbf{P}, \mathbf{P'} \)符号。

因为\( f \)为\( I \)上的分段常数函数,因此\( \exists I \)的分区\( \mathbf{P} \)使得 \( f \)针对\( \mathbf{P} \)为分段常数函数。又\( \{ J, K \} \)也是\( I \)的分区,因此根据引理11.1.18,可得\( \mathbf{P'} := \mathbf{P} \# \{ J, K \} \)也是\( I \)的分区且比\( \mathbf{P}, \{ J, K \} \)都精细。再根据引理11.2.7,可得\( f \)针对\( \mathbf{P'} \)为分段常数函数。

令\( \mathbf{P_J} := \{ L \cap J : L \in \mathbf{P} \} \),我们证明下\( \mathbf{P_J} \)是\( J \)的分区: \( \forall x \in J \),有\( x \in I \),又\( \mathbf{P} \)是\( I \)的分区,因此\( \exists L_x \in \mathbf{P}, x \in L_x \),加上\( x \in J \),可得\( x \in L_x \cap J \),进而\( x \in \mathbf{P_J} \)。接着证明\( \mathbf{P_J} \)中的区间互不相交,假设\( \mathbf{P_J} \)中存在相交的的区间,即\( \exists L_1, L_2 \in \mathbf{P}, (L_1 \cap J) \cap (L_2 \cap J) \neq \emptyset \),这意味着\( \exists x \)满足\( x \in L_1 \)且\( x \in L_2 \),这和\( \mathbf{P} \)为\( I \)的区间矛盾,因此假设不成立,有\( \mathbf{P_J} \)中的区间互不相交。至此,我们证明了\( \mathbf{P_J} \)是\( J \)的分区。同理可证\( \mathbf{P_K} := \{ L \cap K : L \in \mathbf{P} \} \),是\( K \)的分区。

由\( \mathbf{P_J} \subseteq \mathbf{P'} \)以及\( f \)针对\( \mathbf{P'} \)为分段常数函数,可得\( f \)在\( \mathbf{P_J} \)的区间上的函数值均为常数值,而\( f_J \)在这些区间上的函数值和\( f \)相同,因此\( f_J \)也在\( \mathbf{P_J} \)的区间上的函数值均为常数值,即\( f_J \)为\( \mathbf{P_J} \)上的分段常数函数。同理可证\( f_K \)为\( \mathbf{P_K} \)上的分段常数函数。

注意到\( \mathbf{P'} = \mathbf{P_J} \cup \mathbf{P_K} \)且 \( \mathbf{P_J} \cap \mathbf{P_K} = \emptyset \),可得 \( p.c. \int_{I} f = p.c. \int_{[\mathbf{P'}]} f = \sum_{L \in \mathbf{P'}} c_L|L| = (\sum_{L_J \in \mathbf{P_J}} c_{L_J}|L_J|) + (\sum_{L_K \in \mathbf{P_K}} c_{L_K}|L_K|) = (p.c. \int_{J} f|_J) + (p.c. \int_{K} f|_K) \)。

证毕。

章节11.3

附注

这里将“majorize”翻译成“主要于”,类似的,“minorize”翻译成“次要于”。

练习11.3.1

题目:

Let \( f: I \to \mathbf{R}, g: I \to \mathbf{R} \), and \( h : I \to \mathbf{R} \) be functions. Show that if \( f \) majorizes \( g \) and \( g \) majorizes \( h \), then \( f \) majorizes \( h \). Show that if \( f \) and \( g \) majorize each other, then they must be equal.

证明:

如果\( f \)主要于\( g \)且\( g \)主要于\( h \): \( \forall x \in I \),因为\( f \)主要于\( g \),因此\( f(x) \geq g(x) \),因为\( g \)主要于\( h \),因此\( g(x) \geq h(x) \),于是\( f(x) \geq g(x) \geq h(x) \)。简而言之,\( \forall x \in I \),有\( f(x) \geq h(x) \),即\( f \)主要于\( h \)。

如果\( f \)主要于\( g \)且\( g \)主要于\( f \): \( \forall x \in I \),因为\( f \)主要于\( g \),因此\( f(x) \geq g(x) \),因为\( g \)主要于\( f \),因此\( g(x) \geq f(x) \),可得\( f(x) = g(x) \)。综上,\( f = g \)。

证毕。

练习11.3.2

题目:

Let \( f: I \to \mathbf{R}, g: I \to \mathbf{R} \), and \( h: I \to \mathbf{R} \) be functions. If \( f \) majorizes \( g \), is it true that \( f + h \) majorizes \( g + h \)? Is it true that \( f \cdot h \) majorizes \( g \cdot h \)? If \( c \) is a real number, is it true that \( cf \) majorizes \( cg \)?

证明:

如果\( f \)主要于\( g \),那么\( f + h \)主要于\( g + h \)吗?

答案是肯定的,以下是证明:

如果\( f \)主要于\( g \),则\( \forall x \in I, f(x) \geq g(x) \),进而\( (f + h)(x) = f(x) + h(x) \geq g(x) + h(x) = (g + h)(x) \),可得\( f + h \)主要于\( g + h \)。

如果\( f \)主要于\( g \),那么\( f \cdot h \)主要于\( g \cdot h \)吗?

不一定,举个例子,令\( f \)为常数函数\( 2 \),\( g \)为常数函数\( 1 \), \( h \)为常数函数\( -1 \),则\( f \)主要于\( g \),但是此时\( g \cdot h \)主要于\( f \cdot h \)。

加上额外的条件\( \forall x \in I, h(x) \geq 0 \),才能保证\( f \cdot h \)主要于\( g \cdot h \)的结论。

给定\( c \in \mathbf{R} \),\( cf \)主要于\( cg \)吗?

不一定,举个例子,令\( c := -1 \),令\( f \)为常数函数\( 2 \), \( g \)为常数函数\( 1 \),则\( f \)主要于\( g \),但是此时\( cg \)主要于\( cf \)。

加上额外的条件\( c \geq 0 \),才能保证\( cf \)主要于\( cg \)的结论。

证毕。

练习11.3.3

题目:

Prove Lemma 11.3.7.

Lemma 11.3.7的内容:

Let \( f: I \to \mathbf{R} \) be a piecewise constant function on a bounded interval \( I \). Then \( f \) is Riemann integrable, and \( \int_{I} f = p.c. \int_{I} f \).

证明:

因为\( f \)为\( I \)上的分段常数函数,因此\( \exists I \)的分区\( \mathbf{P} \)使得 \( f \)针对\( \mathbf{P} \)为分段常数函数,令\( M_1 := \min(\{ c_J : J \in \mathbf{P} \}), M_2 := \max(\{ c_J : J \in \mathbf{P} \}) \)(分区中的区间数量是有限的,故\( \max, \min \)是良好定义的),令\( M := \max(|M_1|, |M_2|) \),可得\( \forall x \in I, |f(x)| \leq M \),即\( f \)有界。

令\( U := \{ p.c. \int_{I} g : g \text{为} I \text{上主要于} f \text{的分段常数函数} \} \),令\( L := \{ p.c. \int_{I} g : g \text{为} I \text{上次要于} f \text{的分段常数函数} \} \)。

由于\( f \)主要于\( f \),因此\( f \in U \),进而\( \overline{\int}_{I} f = \inf(U) \leq p.c. \int_{I} f \)。由于\( f \)次要于\( f \),因此\( f \in L \),进而\( \underline{\int}_{I} f = \sup(L) \geq p.c. \int_{I} f \)。综上,\( \overline{\int}_{I} f \leq \underline{\int}_{I} f \),加上由引理11.3.3,可得\( \underline{\int}_{I} f \leq \overline{\int}_{I} f \),于是可得\( \underline{\int}_{I} f = \overline{\int}_{I} f \),即\( f \)黎曼可积。

\( \forall g \in U \),有\( g \)主要于\( f \),因此\( \forall x \in I, g(x) \geq f(x) \),进而根据定理11.2.16的5,可得\( p.c. \int_{I} g \geq p.c. \int_{I} f \),综上可得,\( p.c. \int_{I} f \)为\( U \)的下界,于是有\( \overline{\int}_{I} f = \inf(U) \geq p.c. \int_{I} f \)。 \( \forall g \in L \),有\( g \)次要于\( f \),因此\( \forall x \in I, g(x) \leq f(x) \),进而根据定理11.2.16的5,可得\( p.c. \int_{I} g \leq p.c. \int_{I} f \),综上可得,\( p.c. \int_{I} f \)为\( L \)的上界,于是有\( \underline{\int}_{I} f = \sup(L) \leq p.c. \int_{I} f \)。而\( \int_{I} f = \underline{\int}_{I} f = \overline{\int}_{I} f \),可得\( \int_{I} f \leq p.c. \int_{I} f \leq \int_{I} f \),进而可得\( \int_{I} f = p.c. \int_{I} f \)。

证毕。

练习11.3.4

题目:

Prove Lemma 11.3.11.

Lemma 11.3.11的内容:

Let \( f: I \to \mathbf{R} \) be a bounded function on a bounded interval \( I \), and let \( g \) be a function which majorizes \( f \) and which is piecewise constant with respect to some partition \( \mathbf{P} \) of \( I \). Then \( p.c. \int_{I} g \geq U(f, \mathbf{P}) \). Similarly, if \( h \) is a function which minorizes \( f \) and is piecewise constant with respect to \( \mathbf{P} \), then \( p.c. \int_{I} h \leq L(f, \mathbf{P}) \).

证明:

证明如果分段常数函数\( g \)主要于\( f \),则\( p.c. \int_{I} g \geq U(f, \mathbf{P}) \):

如果\( I = \emptyset \),则\( p.c. \int_{I} g = 0 \geq 0 = U(f, \mathbf{P}) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall J \neq \emptyset \in \mathbf{P} \),因为\( g \)主要于\( f \),因此\( \forall x \in J, g(x) \geq f(x) \),进而\( \sup_{x \in J} g(x) \geq \sup_{x \in J} f(x) \),而\( \forall x \in J, g(x) \)均\( = g \)在\( J \)上的常数值\( c_J \),因此\( c_J = \sup_{x \in J} g(x) \geq \sup_{x \in J} f(x) \),进而可得\( c_J|J| \geq (\sup_{x \in J} f(x))|J| \)。综上可得,\( p.c. \int_{I} g = \sum_{J \in \mathbf{P}} c_J|J| = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J|) + (\sum_{J \in \mathbf{P} : J = \emptyset} c_J|J|) = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J|) + 0 = \sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J| \geq \sum_{J \in \mathbf{P} : J \neq \emptyset} (\sup_{x \in J} f(x))|J| \),即\( p.c. \int_{I} g \geq U(f, \mathbf{P}) \)。

证明如果分段常数函数\( h \)次要于\( f \),则\( p.c. \int_{I} h \leq L(f, \mathbf{P}) \):

把“g”改成“h”,“主要”改成“次要”,“\( \geq \)”改成“\( \leq \)”,“\( \sup \)”改成“\( \inf \)”,“U”改成“L”就行,如下:

如果\( I = \emptyset \),则\( p.c. \int_{I} h = 0 \leq 0 = L(f, \mathbf{P}) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall J \neq \emptyset \in \mathbf{P} \),因为\( h \)次要于\( f \),因此\( \forall x \in J, h(x) \leq f(x) \),进而\( \inf_{x \in J} h(x) \leq \inf_{x \in J} f(x) \),而\( \forall x \in J, h(x) \)均\( = h \)在\( J \)上的常数值\( c_J \),因此\( c_J = \inf_{x \in J} h(x) \leq \inf_{x \in J} f(x) \),进而可得\( c_J|J| \leq (\inf_{x \in J} f(x))|J| \)。综上可得,\( p.c. \int_{I} h = \sum_{J \in \mathbf{P}} c_J|J| = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J|) + (\sum_{J \in \mathbf{P} : J = \emptyset} c_J|J|) = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J|) + 0 = \sum_{J \in \mathbf{P} : J \neq \emptyset} c_J|J| \leq \sum_{J \in \mathbf{P} : J \neq \emptyset} (\inf_{x \in J} f(x))|J| \),即\( p.c. \int_{I} h \leq L(f, \mathbf{P}) \)。

证毕。

练习11.3.5

题目:

Prove Proposition 11.3.12. (Hint: you will need Lemma 11.3.11, even though this Lemma will only do half of the job.)

Proposition 11.3.12的内容:

Let \( f: I \to \mathbf{R} \) be a bounded function on a bounded interval \( I \). Then \( \overline{\int}_{I} f = \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text { is a partition of } I \}) \) and \( \underline{\int}_{I} f = \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text { is a partition of } I \}) \).

证明:

证明\( \overline{\int}_{I} f = \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \):

如果\( I = \emptyset \),则\( \overline{\int}_{I} f = 0 = \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall g \in \{ p.c. \int_{I} g : g \text{为} I \text{上主要于} f \text{的分段常数函数} \} \),有\( g \)主要于\( f \),\( \forall I \)的分区\( \mathbf{P} \),根据引理11.3.11,可得\( p.c. \int_{I} g \geq U(f, \mathbf{P}) \),于是可得\( p.c. \int_{I} g \)为 \( \{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \} \)的上界,进而可得\( p.c. \int_{I} g \geq \sup(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \geq \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \) (重点在于\( p.c. \int_{I} g \geq \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \))。综上,可得\( \overline{\int}_{I} f = \inf(\{ p.c. \int_{I} g : g \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \geq \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

还差证明\( \overline{\int}_{I} f \leq \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \),假设\( \overline{\int}_{I} f > \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \),则\( \exists I \)的分区\( \mathbf{P'} \),使得\( U(f, \mathbf{P'}) < \overline{\int}_{I} f \) (1) ,定义函数\( h: I \to \mathbf{R}, \forall y \in I \),因为\( \mathbf{P'} \)为\( I \)的分区,因此\( \exists \)唯一的\( J_y \in \mathbf{P'}, y \in J_y \)(注意,这意味着\( J_y \)不为空),令\( h(y) := \sup_{x \in J_y} f(x) \),易得\( h \)为\( I \)上的分段常数函数(属于同一个区间\( J \)内的点的函数值均为\( \sup_{x \in J} f(x) \))且\( p.c. \int_{I} h = p.c. \int_{[\mathbf{P'}]} h = \sum_{J \in \mathbf{P'} : J \neq \emptyset} (\sup_{x \in J} f(x))|J| = U(f, \mathbf{P'}) \) (2) ,又\( \forall x \in I, h(x) \geq f(x) \),因此也有\( h \)主要于\( f \),于是可得\( p.c. \int_{I} h \in \{ p.c. \int_{I} g : g \text{为} I \text{上主要于} f \text{的分段常数函数} \} \),进而可得\( \overline{\int}_{I} f = \inf(\{ p.c. \int_{I} g : g \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \leq p.c. \int_{I} h \),加上(2),可得\( \overline{\int}_{I} f \leq U(f, \mathbf{P'}) \),这和(1)矛盾,因此假设不成立,有\( \overline{\int}_{I} f \leq \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

综上,可得\( \overline{\int}_{I} f = \inf(\{ U(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

证明\( \underline{\int}_{I} f = \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \):

和上面差不多,简单点就是写易证。

如果\( I = \emptyset \),则\( \underline{\int}_{I} f = 0 = \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall g \in \{ p.c. \int_{I} g : g \text{为} I \text{上次要于} f \text{的分段常数函数} \} \),有\( g \)次要于\( f \),\( \forall I \)的分区\( \mathbf{P} \),根据引理11.3.11,可得\( p.c. \int_{I} g \leq L(f, \mathbf{P}) \),于是可得\( p.c. \int_{I} g \)为 \( \{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \} \)的下界,进而可得\( p.c. \int_{I} g \leq \inf(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \leq \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。综上,可得\( \underline{\int}_{I} f = \sup(\{ p.c. \int_{I} g : g \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \leq \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

还差证明\( \underline{\int}_{I} f \geq \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \),假设\( \underline{\int}_{I} f < \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \),则\( \exists I \)的分区\( \mathbf{P'} \),使得\( L(f, \mathbf{P'}) > \underline{\int}_{I} f \) (1) ,定义函数\( h: I \to \mathbf{R}, \forall y \in I \),因为\( \mathbf{P'} \)为\( I \)的分区,因此\( \exists \)唯一的\( J_y \in \mathbf{P'}, y \in J_y \)(注意,这意味着\( J_y \)不为空),令\( h(y) := \inf_{x \in J_y} f(x) \),易得\( h \)为\( I \)上的分段常数函数且\( p.c. \int_{I} h = p.c. \int_{[\mathbf{P'}]} h = \sum_{J \in \mathbf{P'} : J \neq \emptyset} (\inf_{x \in J} f(x))|J| = L(f, \mathbf{P'}) \) (2) ,又\( \forall x \in I, h(x) \leq f(x) \),因此也有\( h \)次要于\( f \),于是可得\( p.c. \int_{I} h \in \{ p.c. \int_{I} g : g \text{为} I \text{上次要于} f \text{的分段常数函数} \} \),进而可得\( \underline{\int}_{I} f = \sup(\{ p.c. \int_{I} g : g \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \geq p.c. \int_{I} h \),加上(2),可得\( \underline{\int}_{I} f \geq L(f, \mathbf{P'}) \),这和(1)矛盾,因此假设不成立,有\( \underline{\int}_{I} f \geq \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

综上,可得\( \underline{\int}_{I} f = \sup(\{ L(f, \mathbf{P}) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

证毕。

章节11.4

练习11.4.1

题目:

Prove Theorem 11.4.1. (Hint: you may find Theorem 11.2.16 to be useful. For part 2: First do the case \( c > 0 \). Then do the case \( c = −1 \) and \( c = 0 \) separately. Using these cases, deduce the case of \( c < 0 \). You can use earlier parts of the theorem to prove later ones.)

Theorem 11.4.1的内容:

(Laws of Riemann integration). Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) be Riemann integrable functions on \( I \).

  1. The function \( f + g \) is Riemann integrable, and we have \( \int_{I} (f + g) = \int_{I} f + \int_{I} g \).
  2. For any real number \( c \), the function \( cf \) is Riemann integrable, and \( \int_{I} (cf) = c(\int_{I} f) \).
  3. The function \( f - g \) is Riemann integrable, and we have \( \int_{I} (f - g) = \int_{I} f - \int_{I} g \).
  4. If \( f(x) \geq 0 \) for all \( x \in I \), then \( \int_{I} f \geq 0 \).
  5. If \( f(x) \geq g(x) \) for all \( x \in I \), then \( \int_{I} f \geq \int_{I} g \).
  6. If \( f \) is the constant function \( f(x) = c \) for all \( x \) in \( I \), then \( \int_{I} f = c|I| \).
  7. Let \( J \) be a bounded interval containing \( I \) (i.e. \( I \subseteq J \)), and let \( F: J \to \mathbf{R} \) be the function \( F(x) := \begin{cases} f(x) & \text{ if } x \in I \\ 0 & \text{ if } x \notin I \end{cases} \), Then \( F \) is Riemann integrable on \( J \), and \( \int_{J} F = \int_{I} f \).
  8. Suppose that \( \{ J, K \} \) is a partition of \( I \) into two intervals \( J \) and \( K \). Then the functions \( f|_J: J \to \mathbf{R} \) and \( f|_K: K \to \mathbf{R} \) are Riemann integrable on \( J \) and \( K \) respectively, and we have \( \int_{I} f = \int_{J} f|_J + \int_{K} f|_K \).

证明:

证明1:

因为\( f, g \)有界,因此\( \exists M_1 > 0, M_2 > 0 \),\( \forall x \in I, |f(x)| \leq M_1, |g(x)| \leq M_2 \),可得\( |(f + g)(x)| = |f(x) + g(x)| \leq M_1 + M_2 \),即\( f + g \)有界。

\( \forall \epsilon > 0 \):

因为\( (\int_{I} f) + \dfrac{\epsilon}{4} > \int_{I} f = \overline{\int}_{I} f = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上主要于\( f \)的分段常数函数\( f_u \)使得\( \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{4} \) (注:根据引理11.3.7,\( p.c. \int_{I} f_u = \int_{I} f_u \))。

因为\( (\int_{I} f) - \dfrac{\epsilon}{4} < \int_{I} f = \underline{\int}_{I} f = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上次要于\( f \)的分段常数函数\( f_l \)使得\( \int_{I} f_l > (\int_{I} f) - \dfrac{\epsilon}{4} \)。

由\( f_u \)主要于\( f \),\( f_l \)次要于\( f \)以及\( f \)黎曼可积,可得 \( (\int_{I} f) - \dfrac{\epsilon}{4} < \int_{I} f_l \leq \underline{\int}_{I} f = \overline{\int}_{I} f \leq \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{4} \) (1) (注:根据引理11.3.11、定理11.3.12、\( f \)黎曼可积以及\( f_u \)主要于\( f \),易得\( \int_{I} f_u \geq \int_{I} f \), \( f_l \)同理)。

同理可得\( \exists I \)上主要于\( g \)的分段常数函数\( g_u \),\( \exists I \)上次要于\( g \)的分段常数函数\( g_l \),使得\( (\int_{I} g) - \dfrac{\epsilon}{4} < \int_{I} g_l \leq \underline{\int}_{I} g = \overline{\int}_{I} g \leq \int_{I} g_u < (\int_{I} g) + \dfrac{\epsilon}{4} \) (2)

由(1)和(2),可得\( (\int_{I} f) + (\int_{I} g) - \dfrac{\epsilon}{2} < (\int_{I} f_l) + (\int_{I} g_l) \leq (\int_{I} f_u) + (\int_{I} g_u) < (\int_{I} f) + (\int_{I} g) + \dfrac{\epsilon}{2} \) (3)

根据引理11.3.7和定理11.2.16的1,可得\( (\int_{I} f_l) + (\int_{I} g_l) = \int_{I} (f_l + g_l) \) 以及\( (\int_{I} f_u) + (\int_{I} g_u) = \int_{I} (f_u + g_u) \),于是(3)变成 \( (\int_{I} f) + (\int_{I} g) - \dfrac{\epsilon}{2} < \int_{I} (f_l + g_l) \leq \int_{I} (f_u + g_u) < (\int_{I} f) + (\int_{I} g) + \dfrac{\epsilon}{2} \) (4)

易证\( f_l + g_l \)次要于\( f + g \),\( f_u + g_u \)主要于\( f - g \),因此根据引理11.3.3、引理11.3.11和定理11.3.12,可得\( \int_{I} (f_l + g_l) \leq \underline{\int}_{I} (f + g) \leq \overline{\int}_{I} (f + g) \leq \int_{I} (f_u + g_u) \),此时根据(4),可得\( (\int_{I} f) + (\int_{I} g) - \dfrac{\epsilon}{2} < \int_{I} (f_l + g_l) \leq \underline{\int}_{I} (f + g) \leq \overline{\int}_{I} (f + g) \leq \int_{I} (f_u + g_u) < (\int_{I} f) + (\int_{I} g) + \dfrac{\epsilon}{2} \) (5) ,进而可得\( (\overline{\int}_{I} (f + g)) - (\underline{\int}_{I} (f + g)) < ((\int_{I} f) + (\int_{I} g) + \dfrac{\epsilon}{2}) - ((\int_{I} f) + (\int_{I} g) - \dfrac{\epsilon}{2}) = \epsilon \),由于该不等式对任意\( \epsilon \)都成立,因此可得\( \overline{\int}_{I} (f + g) = \underline{\int}_{I} (f + g) \),即\( f + g \)黎曼可积。根据\( f + g \)黎曼可积和(5),可得\( (\int_{I} f) + (\int_{I} g) - \dfrac{\epsilon}{2} < \int_{I} (f + g) < (\int_{I} f) + (\int_{I} g) + \dfrac{\epsilon}{2} \),可得\( |(\int_{I} (f + g)) - ((\int_{I} f) + (\int_{I} g))| < \dfrac{\epsilon}{2} < \epsilon \),由于该不等式对任意\( \epsilon \)都成立,因此可得 \( \int_{I} (f + g) = (\int_{I} f) + (\int_{I} g) \)。

证明2:

因为\( f \)有界,因此\( \exists M > 0 \),\( \forall x \in I, |f(x)| \leq M \),可得\( |(cf)(x)| = |cf(x)| \leq cM \),即\( cf \)有界。

先考虑\( c > 0 \)的情况:

\( \forall \epsilon > 0 \):

因为\( (\int_{I} f) + \dfrac{\epsilon}{2c} > \int_{I} f = \overline{\int}_{I} f = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上主要于\( f \)的分段常数函数\( f_u \)使得\( \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2c} \)。

因为\( (\int_{I} f) - \dfrac{\epsilon}{2c} < \int_{I} f = \underline{\int}_{I} f = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上次要于\( f \)的分段常数函数\( f_l \)使得\( \int_{I} f_l > (\int_{I} f) - \dfrac{\epsilon}{2c} \)。

由\( f_u \)主要于\( f \),\( f_l \)次要于\( f \)以及\( f \)黎曼可积,可得 \( (\int_{I} f) - \dfrac{\epsilon}{2c} < \int_{I} f_l \leq \underline{\int}_{I} f = \overline{\int}_{I} f \leq \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2c} \),进而可得\( c(\int_{I} f) - \dfrac{\epsilon}{2} < c(\int_{I} f_l) \leq c(\int_{I} f_u) < c(\int_{I} f) + \dfrac{\epsilon}{2} \) (1)

根据引理11.3.7和定理11.2.16的2,可得\( c(\int_{I} f_l) = \int_{I} (cf_l) \) 以及\( c(\int_{I} f_u) = \int_{I} (cf_u) \),于是(1)变成 \( c(\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} (cf_l) \leq \int_{I} (cf_u) < c(\int_{I} f) + \dfrac{\epsilon}{2} \) (2)

易证\( cf_l \)次要于\( cf \),\( cf_u \)主要于\( cf \),因此根据引理11.3.3、引理11.3.11和定理11.3.12,可得\( \int_{I} (cf_l) \leq \underline{\int}_{I} (cf) \leq \overline{\int}_{I} (cf) \leq \int_{I} (cf_u) \),此时根据(2),可得\( c(\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} (cf_l) \leq \underline{\int}_{I} (cf) \leq \overline{\int}_{I} (cf) \leq \int_{I} (cf_u) < c(\int_{I} f) + \dfrac{\epsilon}{2} \) (3) ,进而可得\( \overline{\int}_{I} (cf) - \underline{\int}_{I} (cf) < (c(\int_{I} f) + \dfrac{\epsilon}{2}) - (c(\int_{I} f) - \dfrac{\epsilon}{2}) = \epsilon \),由于该不等式对任意\( \epsilon \)都成立,因此可得\( \overline{\int}_{I} (cf) = \underline{\int}_{I} (cf) \),即\( cf \)黎曼可积。根据\( cf \)黎曼可积和(3),可得\( c(\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} (cf) < c(\int_{I} f) + \dfrac{\epsilon}{2} \),可得\( |\int_{I} (cf) - c(\int_{I} f)| < \dfrac{\epsilon}{2} < \epsilon \),由于该不等式对任意\( \epsilon \)都成立,因此可得 \( \int_{I} (cf) = c(\int_{I} f) \)。

接着考虑\( c = 0 \)的情况:

此时\( cf \)为常数函数\( 0 \),特别的,它也是\( I \)上的分段常数函数,于是根据引理11.3.7和定理11.2.16的2,可得\( cf \)黎曼可积且\( \int_{I} (cf) = c(\int_{I} f) \)。

最后考虑\( c < 0 \)的情况,此时\( -c > 0 \):

由定理11.2.16的2,可得 \( -\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \} = \{ -(p.c. \int_{I} h) : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \} = \{ p.c. \int_{I} (-h) : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \} = \{ p.c. \int_{I} (-h) : -h \text{为} I \text{上主要于} cf \text{的分段常数函数} \} = \{ p.c. \int_{I} h : h \text{为} I \text{上主要于} cf \text{的分段常数函数} \} \) (4) ,同理可得, \( -\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \} = \{ -(p.c. \int_{I} h) : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \} = \{ p.c. \int_{I} (-h) : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \} = \{ p.c. \int_{I} (-h) : -h \text{为} I \text{上次要于} cf \text{的分段常数函数} \} = \{ p.c. \int_{I} h : h \text{为} I \text{上次要于} cf \text{的分段常数函数} \} \)。根据练习5.5.1和(4),可得\( -\sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \}) = \inf(-\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \}) = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} cf \text{的分段常数函数} \}) \) (5) ,同理可得,\( -\inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \}) = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} cf \text{的分段常数函数} \}) \) (6)

根据之前的证明,有\( -cf \)黎曼可积,可得\( \underline{\int}_{I} (-cf) = \overline{\int}_{I} (-cf) \),这意味着\( \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \}) = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \}) \),进而\( -\sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} -cf \text{的分段常数函数} \}) = -\inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \}) \),根据(5)和(6),可得\( \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} cf \text{的分段常数函数} \}) = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} cf \text{的分段常数函数} \}) \),即\( \overline{\int}_{I} (cf) = \underline{\int}_{I} (cf) \),也就是\( cf \)黎曼可积。

再次根据之前的证明,有\( \int_{I} (-cf) = (-c)(\int_{I} f) \),可得\( -(\int_{I} (-cf)) = c(\int_{I} f) \),根据(6),左边的\( -(\int_{I} (-cf)) = -\inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} -cf \text{的分段常数函数} \}) = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} cf \text{的分段常数函数} \}) = \underline{\int}_{I} (cf) \),而\( \underline{\int}_{I} (cf) = \int_{I} (cf) \),于是可得\( \int_{I} (cf) = c(\int_{I} f) \)。

证明3:

根据2,可得\( -g \)黎曼可积且\( \int_{I} (-g) = -(\int_{I} g) \),再根据1,可得\( f - g = f + (-g) \)黎曼可积且\( \int_{I} (f - g) = \int_{I} (f + (-g)) = (\int_{I} f) + (\int_{I} (-g)) = (\int_{I} f) + (-(\int_{I} g)) = (\int_{I} f) - (\int_{I} g) \)。

证明4:

如果\( \forall x \in I, f(x) \geq 0 \):令\( L := \{ p.c. \int_{I} h : h \text{为} I \text{上次要于} f \text{的分段常数函数} \} \),因为\( f \)黎曼可积,因此\( \int_{I} f = \sup(L) \),假设\( \int_{I} f = \sup(L) < 0 \),明显有常数函数\( 0 \)为\( I \)上次要于\( f \)的分段常数函数,即常数函数\( 0 \)属于\( L \),但常数函数\( 0 > \sup(L) \),这和\( \sup(L) \)为\( L \)的上界矛盾,因此假设不成立,有\( \int_{I} f \geq 0 \)。

证明5:

如果\( \forall x \in I, f(x) \geq g(x) \),则\( (f - g)(x) = f(x) - g(x) \geq 0 \),根据4,可得\( \int_{I} (f - g) \geq 0 \),再根据3,可得\( (\int_{I} f) - (\int_{I} g) = \int_{I} (f - g) \geq 0 \),进而可得\( \int_{I} f \geq \int_{I} g \)。

证明6:

令\( h \)为常数函数\( c \),可得\( \int_{I} h = c|I| \),由\( \forall x \in I, f(x) \geq c \)以及5,可得\( \int_{I} f \geq c|I| \),再次由\( \forall x \in I, f(x) \leq c \)以及5,可得\( \int_{I} f \leq c|I| \),综合可得,\( \int_{I} f = c|I| \)。

证明7:

\( f \)有界,明显\( F \)也是有界的。

\( \forall \epsilon > 0 \):

因为\( (\int_{I} f) + \dfrac{\epsilon}{2} > \int_{f} = \overline{\int}_{I} f = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上主要于\( f \)的分段常数函数\( f_u \)使得\( \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2} \)。定义函数\( F_u: J \to \mathbf{R}, \forall x \in J \),如果\( x \in I \),则令\( F_u(x) := f_u(x) \),如果\( x \notin I \),则令\( F_u(x) := 0 \)。易得\( F_u \)主要于\( F \)。

因为\( (\int_{I} f) - \dfrac{\epsilon}{2} < \int_{f} = \underline{\int}_{I} f = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上次要于\( f \)的分段常数函数\( f_l \)使得\( \int_{I} f_l > (\int_{I} f) - \dfrac{\epsilon}{2} \)。定义函数\( F_l: J \to \mathbf{R}, \forall x \in J \),如果\( x \in I \),则令\( F_l(x) := f_l(x) \),如果\( x \notin I \),则令\( F_l(x) := 0 \)。易得\( F_l \)次要于\( F \)。

由\( F_u \)主要于\( F \),\( F_l \)次要于\( F \)以及定理11.2.16的7,可得\( (\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} f_l = \int_{J} F_l \leq \underline{\int}_{J} F \leq \overline{\int}_{J} F \leq \int_{J} F_u = \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2} \),进一步可得\( |(\overline{\int}_{J} F) - (\underline{\int}_{J} F)| < |((\int_{I} f) + \dfrac{\epsilon}{2}) - ((\int_{I} f) - \dfrac{\epsilon}{2})| = \epsilon \) (1) ,由于该不等式对任意\( \epsilon \)都成立,可得\( \overline{\int}_{J} F = \underline{\int}_{J} F \),即\( F \)黎曼可积。根据\( F \)黎曼可积以及(1),可得\( |(\int_{J} F) - \int_{I} f| < \dfrac{\epsilon}{2} < \epsilon \) 由于该不等式对任意\( \epsilon \)都成立,因此可得 \( \int_{J} F = \int_{I} f \)。

证明8:

\( f \)有界,因此\( f|_J, f|_K \)也有界。

\( \forall \epsilon > 0 \):

因为\( (\int_{I} f) + \dfrac{\epsilon}{2} > \int_{I} f = \inf(\{ p.c. \int_{I} h : h \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上主要于\( f \)的分段常数函数\( f|_u \) 使得\( \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2} \) (1) ,易得\( f_u|_J, f_u|_K \)分别主要于\( f|_J, f|_K \) (2) 。由定理11.2.16的8以及引理11.3.7,可得\( f_u|_J, f_u|_K \)分别在\( J, K \)上为分段常数函数且 \( \int_{I} f_u = (\int_{J} f_u|_J) + (\int_{K} f_u|_K) \) (3)

因为\( (\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} f = \sup(\{ p.c. \int_{I} h : h \text{为} I \text{上次要于} f \text{的分段常数函数} \}) \),因此\( \exists I \)上次要于\( f \)的分段常数函数\( f|_l \) 使得\( \int_{I} f_l > (\int_{I} f) - \dfrac{\epsilon}{2} \) (4) 。易得\( f_l|_J, f_l|_K \)分别次要于\( f|_J, f|_K \) (5) 。由定理11.2.16的8以及引理11.3.7,可得\( f_l|_J, f_l|_K \)分别在\( J, K \)上为分段常数函数且 \( \int_{I} f_l = (\int_{J} f_l|_J) + (\int_{K} f_l|_K) \) (6)

易得\( f_u \)主要于\( f_l \),再由(1)和(4)(还有定理11.2.16的5和引理11.3.7),可得\( (\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} f_l \leq \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2} \) (7) ,特别的,可得\( |\int_{I} f_u - \int_{I} f_l| < |((\int_{I} f) - \dfrac{\epsilon}{2}) - (\int_{I} f) + \dfrac{\epsilon}{2}| = \epsilon \),加上(3)和(6),可得\( |((\int_{J} f_u|_J) + (\int_{K} f_u|_K)) - ((\int_{J} f_l|_J) + (\int_{K} f_l|_K))| = |((\int_{J} f_u|_J) - (\int_{J} f_l|_J)) + ((\int_{K} f_u|_K) - (\int_{K} f_l|_K))| < \epsilon \) (8) ,由(2)和(5),可得\( f_u|_J, f_u|_K \)分别主要于\( f_l|_J, f_l|_K \),于是有 \( (\int_{J} f_u|_J) - (\int_{J} f_l|_J) \geq 0, (\int_{K} f_u|_K) - (\int_{K} f_l|_K) \geq 0 \),结合(8),可得\( 0 \leq ((\int_{J} f_u|_J) - (\int_{J} f_l|_J)) + ((\int_{K} f_u|_K) - (\int_{K} f_l|_K)) < \epsilon \),进而可得\( 0 \leq (\int_{J} f_u|_J) - (\int_{J} f_l|_J) < \epsilon \) (9) 以及\( 0 \leq (\int_{K} f_u|_K) - (\int_{K} f_l|_K) < \epsilon \),而\( \int_{J} f_l|_J \leq \underline{\int}_{J} f|_J \leq \overline{\int}_{J} f|_J \leq \int_{J} f_u|_J \),于是结合(9),可得\( |(\overline{\int}_{J} f|_J) - (\underline{\int}_{J} f|_J)| \leq |(\int_{J} f_u|_J) - (\int_{J} f_l|_J)| = (\int_{J} f_u|_J) - (\int_{J} f_l|_J) < \epsilon \),由于该不等式对所有\( \epsilon \)都成立,因此\( \overline{\int}_{J} f|_J = \underline{\int}_{J} f|_J \),即\( f|_J \)黎曼可积,同理可得\( f|_K \)黎曼可积。

由(2)的\( f_u|_J \)主要于\( f|_J \)、\( f|_J \)黎曼可积以及5(注意,5不带括号),可得,\( \int_{J} f|_J \leq \int_{J} f_u|_J \),同理可得,\( \int_{K} f|_K \leq \int_{K} f_u|K \),于是有\( (\int_{J} f|_J) + (\int_{K} f|_K) \leq (\int_{J} f_u|_J) + (\int_{K} f_u|K) \),类似的,易得\( (\int_{J} f_l|_J) + (\int_{K} f_l|K) \leq (\int_{J} f|_J) + (\int_{K} f|_K) \),加上(3)、(6)和(7),可得\( (\int_{I} f) - \dfrac{\epsilon}{2} < \int_{I} f_l = (\int_{J} f_l|_J) + (\int_{K} f_l|K) \leq (\int_{J} f|_J) + (\int_{K} f|_K) \leq (\int_{J} f_u|_J) + (\int_{K} f_u|K) = \int_{I} f_u < (\int_{I} f) + \dfrac{\epsilon}{2} \),特别的,可得\( |((\int_{J} f_u|_J) + (\int_{K} f_u|K)) - (\int_{I} f)| < \dfrac{\epsilon}{2} < \epsilon \),由于该不等式对所有\( \epsilon \)都成立,因此\( (\int_{J} f_u|_J) + (\int_{K} f_u|K) = \int_{I} f \)。

证毕。

练习11.4.2

题目:

Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a continuous, non-negative function (so \( f(x) ≥ 0 \) for all \( x \in [a, b] \)). Suppose that \( \int_{[a,b]} f = 0 \). Show that \( f(x) = 0 \) for all \( x \in [a, b] \). (Hint: argue by contradiction.)

证明:

假设\( \exists c \in [a,b], f( c) \neq 0 \),由\( f \)为非负函数,可得\( f( c) > 0 \),令\( \epsilon := \dfrac{f( c)}{2} \),因为\( f \)为连续函数,因此\( \exists \delta > 0, \forall x \in [a, b], |x - c| \leq \delta, |f(x) - f( c)| \leq \epsilon \),即\( f( c) - \epsilon \leq f(x) \leq f( c) + \epsilon \)。令\( l := \max(a, c - \delta), r := \min(b, c + \delta) \),由于\( a < b, c - \delta < b \)且\( a < c + \delta, c - \delta < c + \delta \),因此有\( l < r \),除此之外,还易得\( \forall x \in [l, r] \),有\( x \in [a, b] \)且\( |x - c| \leq \delta \),进而有\( f( c) - \epsilon \leq f(x) \leq f( c) + \epsilon \)。定义函数\( g: [a, b] \to \mathbf{R}, \forall x \in [a, b] \),如果\( x \in [l, r] \),则令\( g(x) := f( c) - \epsilon \),否则,令\( g(x) := 0 \),明显\( g(x) \)为\( [a, b] \) 上的分段常数函数且\( \{ [a, l), [l, r], (r, b] \} \)为\( [a, b] \)的分区,因此\( \int_{[a, b]} g(x) = 0 \times (l - a) + (f( c) - \epsilon) \times (r - l) + 0 \times (b - r) = (f( c) - \epsilon) \times (r - l) \),易得\( \forall x \in [a, b], g(x) \leq f(x) \),因此根据定理11.4.1的5,可得\( \int_{[a, b]} f \geq \int_{[a, b]} g = (f( c) - \epsilon) \times (r - l) \),这里\( f( c) - \epsilon = f( c) - \dfrac{f( c)}{2} = \dfrac{f( c)}{2} > 0 \),又\( l < r \),因此\( r - l > 0 \),于是可得\( \int_{[a, b]} f \geq (f( c) - \epsilon) \times (r - l) > 0 \),这和\( \int_{[a, b]} f = 0 \)矛盾,因此假设不成立,有\( \forall x \in [a, b], f(x) = 0 \)。

证毕。

练习11.4.3

题目:

Let \( I \) be a bounded interval, let \( f: I \to \mathbf{R} \) be a Riemann integrable function, and let \( \mathbf{P} \) be a partition of \( I \). Show that \( \int_{I} f = \sum_{J \in \mathbf{P}} \int_{J} f \).

证明:

令\( n := \#(\mathbf{P}) \),对\( n \)进行数学归纳。

当\( n = 0 \)时,有\( I = \emptyset \),此时\( \int_{I} f = 0 = \sum_{J \in \mathbf{P}} \int_{J} f \),基础情况成立。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时, \( \mathbf{P} \)中必然有一个最右的区间\( I_r \in \mathbf{P} \),具体的,它满足\( \forall x \in I - I_r, \forall y \in I_r, x \leq I_r \),令\( \mathbf{P'} := \mathbf{P} - \{ I_r \} \),易得\( \mathbf{P'} \)为\( I - I_r \)的分区,除此之外,还可得\( \{ I - I_r, I_r \} \)为\( I \)的分区,根据定理11.4.1的8,可得 \( f|_{I - I_r}, f|_{I_r} \)黎曼可积且\( \int_{I} f = (\int_{I - I_r} f|_{I - I_r}) + (\int_{I_r} f|_{I_r}) = (\int_{I - I_r} f) + (\int_{I_r} f) \),由于\( \#(\mathbf{P'}) = k \),根据归纳假设,可得\( \int_{I - I_r} f = \sum_{J \in \mathbf{P'}} \int_{J} f \),于是可得,\( \int_{I} f = (\int_{I - I_r} f) + (\int_{I_r} f) = (\sum_{J \in \mathbf{P'}} \int_{J} f) + (\int_{I_r} f) = (\sum_{J \in \mathbf{P}} \int_{J} f) \)。

证毕。

练习11.4.4

题目:

Without repeating all the computations in the above proofs, give a short explanation as to why the remaining cases of Theorem 11.4.3 and Theorem 11.4.5 follow automatically from the cases presented in the text. (Hint: from Theorem 11.4.1 we know that if \( f \) is Riemann integrable, then so is \( -f \).)

Theorem 11.4.3的内容:

(Max and min preserve integrability) Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) be a Riemann integrable function. Then the functions \( \max(f, g): I \to \mathbf{R} \) and \( \min(f, g): I \to \mathbf{R} \) defined by \( \max(f, g)(x) := \max(f(x), g(x)) \) and \( \min(f, g)(x) := \min(f(x), g(x)) \) are also Riemann integrable.

Theorem 11.4.5的内容:

(Products preserve Riemann integrability) Let \( I \) be a bounded interval. If \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) are Riemann integrable, then \( fg: I \to \mathbf{R} \) is also Riemann integrable.

解释:

定理11.4.3:

注意到\( \min(f, g) = -\max(-f, -g) \),根据定理11.4.1的2,可得\( -f, -g \)均黎曼可积,再根据定理11.4.3的证明中对\( \max \)情况的证明,可得\( \max(-f, -g) \)黎曼可积,最后再次根据定理11.4.1的2,可得\( \min(f, g) = -\max(-f, -g) \)黎曼可积。

定理11.4.5:

注意到\( f_+g_- = f_+(-g)_+, f_-g_+ = (-f)_+g_+, f_-g_- = (-f)_+(-g)_+ \),根据定理11.4.1的2,可得\( -g \)黎曼可积,于是将\( f, -g \)代入定理11.4.5的\( f, g \),再根据定理11.4.5的证明,可得\( f_+(-g)_+ \)黎曼可积,其他两种情况同理。

章节11.5

练习11.5.1

题目:

Prove Proposition 11.5.6. (Hint: use Theorem 11.4.1(1) and (7).)

Proposition 11.5.6的内容:

Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) be both piecewise continuous and bounded. Then \( f \) is Riemann integrable.

证明:

因为\( f \)在\( I \)上分段连续,因此\( \exists I \)的分区\( \mathbf{P} \)使得 \( \forall J \in \mathbf{P}, f|_J \)在\( J \)上连续,又\( f \)有界,因此 \( f|_J \)也有界,根据定理11.5.3,可得\( f|_J \)黎曼可积,定义函数\( F_J: I \to \mathbf{R} \), \( F_J := \begin{cases} f|_J(x) & \text{ if } x \in J \\ 0 & \text{ if } x \notin J \end{cases} \),根据定理11.4.1的7,可得\( F_J \)黎曼可积且\( \int_{I} F_J = \int_{J} f_J \),注意到\( \forall x \in I, f(x) = \sum_{J \in \mathbf{P}} F_J(x) \),即\( f = \sum_{J \in \mathbf{P}} F_J \),反复使用定理11.4.1有限次(因为分区的区间数量是有限的),可得\( f \)黎曼可积。

证毕。

章节11.6

练习11.6.1

题目:

Use Proposition 11.6.1 to prove Corollary 11.6.3. (Hint: adapt the proof of Proposition 11.5.3.)

Corollary 11.6.3的内容:

Let \( I \) be a bounded interval, and let \( f : I \to \mathbf{R} \) be both monotone and bounded. Then \( f \) is Riemann integrable on \( I \).

证明:

如果\( I \)为空集或者仅包含单个点,则\( f \)明显在\( I \)上黎曼可积。

如果\( I \)不为空集且不仅包含单个点,则\( I \)会为\( [a, b], (a, b], (a, b), [a, b) \) 形式的其中一种且\( a < b \)。因为\( f \)有界,因此\( \exists M > 0 \)满足 \( \forall x \in I, -M \leq f(x) \leq M \)。针对\( \forall 0 < \epsilon < \dfrac{b - a}{2} \),有\( [a + \epsilon, b - \epsilon] \subseteq I \),因此\( f|_{[a + \epsilon, b - \epsilon]} \)也为单调函数,于是根据定理11.6.1,可得\( f \)在\( [a + \epsilon, b - \epsilon] \)上黎曼可积,因为\( (\int_{[a + \epsilon, b - \epsilon]} f) + \epsilon > \int_{[a + \epsilon, b - \epsilon]} f = \inf(\{ p.c. \int_{[a + \epsilon, b - \epsilon]} h : h \text{为} [a + \epsilon, b - \epsilon] \text{上主要于} f \text{的分段常数函数} \}) \),因此\( \exists [a + \epsilon, b - \epsilon] \)上主要于\( f \)的分段常数函数\( f_u \)使得 \( \int_{[a + \epsilon, b - \epsilon]} f_u < (\int_{[a + \epsilon, b - \epsilon]} f) + \epsilon \) (1) 。定义函数\( F_u: I \to \mathbf{R}, \forall x \in I \),如果\( x \in [a + \epsilon, b - \epsilon] \),则令\( F_u(x) := f_u(x) \),如果\( x \in I - [a + \epsilon, b - \epsilon] \),则令\( F_u(x) := M \),可得\( F_u \)是\( I \)上的分段常数函数且\( F_u \)主要于\( f \) (注意\( I \)为\( [a, b], (a, b], (a, b), [a, b) \)形式的其中一种),于是根据(1)和定理11.2.16的8、6,可得\( \int_{I} F_u = \epsilon M + (\int_{[a + \epsilon, b - \epsilon]} F_u) + \epsilon M = (\int_{[a + \epsilon, b - \epsilon]} F_u) + 2 \epsilon M = (\int_{[a + \epsilon, b - \epsilon]} f_u) + 2 \epsilon M \leq ((\int_{[a + \epsilon, b - \epsilon]} f) + \epsilon) + 2 \epsilon M = (\int_{[a + \epsilon, b - \epsilon]} f) + (2M + 1)\epsilon \),加上\( F_u \)主要于\( f \),可得\( \overline{\int}_{I} f \leq (\int_{[a + \epsilon, b - \epsilon]} f) + (2M + 1)\epsilon \)。同理可得,\( \underline{\int}_{I} f \geq (\int_{[a + \epsilon, b - \epsilon]} f) - (2M + 1)\epsilon \),于是可得\( |\overline{\int}_{I} f - \underline{\int}_{I} f| \leq (4M + 2)\epsilon \),加上\( \epsilon \)可以任意小,可得\( \overline{\int}_{I} f = \underline{\int}_{I} f \),即\( f \)黎曼可积。

证毕。

练习11.6.2

题目:

Formulate a reasonable notion of a piecewise monotone function, and then show that all bounded piecewise monotone functions are Riemann integrable.

定义(分段单调函数):

令\( I \)为一个有界区间,令\( f: I \to \mathbf{R} \)为一个函数,令\( \mathbf{P} \)为\( I \)的分区,我们称\( f \)针对\( \mathbf{P} \) 为分段单调函数当且仅当 \( \forall J \in \mathbf{P}, f \)在\( J \)上单调(即\( \forall x \geq y \in J, f(x) \geq f(y) \)或者 \( \forall x \leq y \in J, f(x) \leq f(y) \))。特别的,该定义允许分区中不同区间的单调性不同,\( f \)在一个区间可以单调递增,在另外一个区间可以单调递减。

证明有界分段单调函数均黎曼可积:

令\( I \)为一个有界区间,令\( \mathbf{P} \)为\( I \)的分区,令\( f: I \to \mathbf{R} \)为针对\( \mathbf{P} \)分段单调且有界的函数,我们证明\( f \)黎曼可积:

由\( I \)为有界区间、\( f \)针对\( \mathbf{P} \)分段单调以及\( f \)有界,可得\( \forall J \in \mathbf{P}, J \)为有界区间、\( f \)在\( J \)上单调且有界,进而根据推论11.6.3,可得\( f \)在\( J \)上黎曼可积。

\( \forall J \in \mathbf{P} \),定义函数\( F_J: I \to \mathbf{R}, \forall x \in I \),如果\( x \in J \),则令\( F_J(x) := f(x) \),否则令\( F_J(x) := 0 \),易得\( f = \sum_{J \in \mathbf{P}} F_J \),反复使用定理11.4.1有限次(因为分区的区间数量是有限的),可得\( f \)黎曼可积。

证毕。

练习11.6.3

题目:

Prove Proposition 11.6.4. (Hint: what is the relationship between the sum \( \sum_{n = 1}^{N} f(n) \), the sum \( \sum_{n = 0}^{N - 1} f(n) \), and the integral \( \int_{[0, N]} f \) ?)

Proposition 11.6.4的内容:

(Integral test). Let \( f: [0, \infty) \to \mathbf{R} \) be a monotone decreasing function which is non-negative (i.e., \( f(x) \geq 0 \) for all \( x \geq 0 \)). Then the \( \sum_{n = 0}^{\infty} f(n) \) is convergent if and only if \( \sup_{N > 0} \int_{[0, N]} f \) is finite.

证明:

\( \forall N > 0 \),定义 \( \mathbf{P} := \{ [n, n + 1) : n \in \mathbf{N}, 0 \leq n \leq N - 1 \} \cup \{ N \} \),明显\( \mathbf{P} \)是\( [0, N] \)的分区。定义函数\( f_{uN}: [0, N] \to \mathbf{R}, \forall x \in [0, N] \),如果\( x = N \),则令\( f(x) := f(N) \),如果\( x \neq N \),则此时\( \exists \)唯一的\( J = [n, n + 1) \in \mathbf{P} \)使得\( x \in J \),我们令\( f(x) := f(n) \),易得\( f_{uN} \)在\( [0, N] \)上主要于\( f \)。定义函数\( f_{lN}: [0, N] \to \mathbf{R}, \forall x \in [0, N] \),如果\( x = N \),则令\( f(x) := f(N) \),如果\( x \neq N \),则此时\( \exists \)唯一的\( J = [n, n + 1) \in \mathbf{P} \)使得\( x \in J \),我们令\( f(x) := f(n + 1) \)(注:虽然\( n + 1 \)不在区间\( [n, n + 1) \)里面,但是这并不妨碍我们取\( f(n + 1) \)作为\( x \in [n, n + 1) \)的函数值),易得\( f_{lN} \)在\( [0, N] \)上次要于\( f \)。根据定理11.6.1,可得\( f \)在\( [0, N] \)上黎曼可积,因此\( \int_{[0, N]} f \)是存在的,又\( f_{uN} \)在\( [0, N] \)上主要于\( f \)、\( f_{lN} \)在\( [0, N] \)上次要于\( f \),因此可得\( \sum_{n = 1}^{N} f(n) = \int_{[0, N]} f_{lN} \leq \int_{[0, N]} f \leq \int_{[0, N]} f_{uN} = \sum_{n = 0}^{N - 1} f(n) \) (1)

必要性:

如果\( \sum_{n = 0}^{\infty} f(n) \)收敛:假设\( \sup_{N > 0} \int_{[0, N]} f \)非有限,则由于\( \forall x \geq 0, f(x) \geq 0 \),可得\( \sup_{N > 0} \int_{[0, N]} f = +\infty \),于是有\( \forall M \in \mathbf{R} \),均\( \exists N > 0, \int_{[0, N]} f > M \),此时因为\( f(N) \geq 0 \),加上(1),可得\( \sum_{n = 0}^{N} f(n) = (\sum_{n = 0}^{N - 1} f(n)) + f(N) \geq \sum_{n = 0}^{N - 1} f(n) \geq \int_{[0, N]} f > M \),这和\( \sum_{n = 0}^{\infty} f(n) = \lim_{N \to \infty} (\sum_{n = 0}^{N} f(n)) \)收敛矛盾(注:收敛必有界),故假设不成立,有\( \sup_{N > 0} \int_{[0, N]} f \)有限。

充分性:

如果\( \sup_{N > 0} \int_{[0, N]} f \)有限:假设\( \sum_{n = 0}^{\infty} f(n) \)不收敛,则由于\( \forall x \geq 0, f(x) \geq 0 \),可得\( \forall M \in \mathbf{R} \),均\( \exists N > 0, \sum_{n = 0}^{N} f > M + f(0) \),可得\( \sum_{n = 1}^{N} f = (\sum_{n = 0}^{N} f) - f(0) > M \),进而由(1)可得,\( \int_{[0, N]} f \geq \sum_{n = 1}^{N} f(n) > M \),这和\( \sup_{N > 0} \int_{[0, N]} f \)有限矛盾,故假设不成立,有\( \sum_{n = 0}^{\infty} f(n) \)不收敛。

证毕。

练习11.6.4

题目:

Give examples to show that both directions of the integral test break down if f is not assumed to be monotone decreasing.

例子1(\( \sum_{n = 0}^{\infty} f(n) \)收敛,但\( \sup_{N > 0} \int_{[0, N]} f \)非有限):

注意到级数仅取整数点,而积分是“连续”进行的,我们可以在两个整数点之间定义一些级数取不到的值,从而做到级数收敛,但是积分非有限。

定义函数\( f: [0, \infty] \to \mathbf{R}, \forall x \geq 0 \),如果\( x \in \mathbf{N} \),则令\( f(x) := 0 \),否则,令\( f(x) := 1 \),可得\( \sum_{n = 0}^{\infty} f(n) = 0 \),但是\( \forall N > 0, \int_{[0, N]} f = N \),进而\( \sup_{N > 0} \int_{[0, N]} f = +\infty \)非有限。

例子2(\( \sup_{N > 0} \int_{[0, N]} f \)有限,但\( \sum_{n = 0}^{\infty} f(n) \)不收敛):

注意到积分对单点的值不敏感(分区中区间长度为0),因此可以定义分段常数函数,在整数点的时候给一些加起来不会收敛的值。

定义函数\( f: [0, \infty] \to \mathbf{R}, \forall x \geq 0 \),如果\( x \in \mathbf{N} \),则令\( f(x) := 1 \),否则,令\( f(x) := 0 \),可得\( \forall N > 0, \int_{[0, N]} f = 0 \),但是\( \sum_{n = 0}^{N} f(n) = N + 1 \),进而\( \sum_{n = 0}^{\infty} f(n) = \lim_{N \to \infty} (\sum_{n = 0}^{N} f(n)) \)不收敛。

练习11.6.5

题目:

Use Proposition 11.6.4 to prove Corollary 11.6.5.

Corollary 11.6.5的内容:

Let \( p \) be a real number. Then \( \sum_{n = 1}^{\infty} \dfrac{1}{n^p} \) converges absolutely when \( p > 1 \) and diverges when \( p \leq 1 \). (For this exercise, you may use the second Fundamental Theorem of Calculus (Theorem 11.9.4); there is no circularity, because Corollary 11.6.5 is not used in the proof of that theorem.)

Theorem 11.9.4的内容:

(Second Fundamental Theorem of Calculus). Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a Riemann integrable function. If \( F: [a, b] \to \mathbf{R} \) is an antiderivative of \( f \), then \( \int_{[a, b]} f = F(b) - F(a) \).

Definition 11.9.3 (Antiderivatives)的内容:

Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) be a function. We say that a function \( F: I \to \mathbf{R} \) is an antiderivative of \( f \) if \( F \) is differentiable on \( I \) and \( F'(x) = f(x) \) for all limit points \( x \) of \( I \).

证明:

\( \sum_{n = 1}^{\infty} \dfrac{1}{n^p} = \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \),根据定理7.2.14的4,可得\( \sum_{n = 1}^{\infty} \dfrac{1}{n^p} \)收敛当且仅当 \( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)收敛,故我们仅需要证明 \( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)收敛或者发散就行。

定义函数\( f: [0, \infty), \forall x \geq 0 \),令\( f(x) := \dfrac{1}{(x + 1)^p} \)。

当\( 0 < p < 1 \)或\( p > 1 \)时,定义函数\( F: [0, +\infty), \forall x \geq 0 \),令\( F(x) := \dfrac{1}{1 - p} (x + 1)^{1 - p} \),根据练习10.4.3的2(注:\( x + 1 > 0 \),不会等于\( 0 \))、定理10.1.15以及定理10.1.13的5,可得\( \forall x \geq 0, F'(x) = \dfrac{1}{(x + 1)^p} = f(x) \),根据定理11.9.4以及(1),可得\( \forall N > 0, \int_{[0, N]} f = F(N) - F(0) = (\dfrac{1}{1 - p} (N + 1)^{1 - p}) - \dfrac{1}{1 - p} = \dfrac{1}{1 - p} ((N + 1)^{1 - p} - 1) \) (1)

当\( p > 1 \)时,\( f \)为单调递减函数且\( \forall x \geq 0, f(x) \geq 0 \)。由\( p > 1 \)以及(1),可得\( \forall N > 0, \int_{[0, N]} f \)的值单调递减,故\( \sup_{N > 0} \int_{[0, N]} f = \int_{[0, 0]} f = \dfrac{1}{1 - p} (1 - 1) = 0 \)有限,进而根据定理11.6.4,可得 \( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)收敛。

当\( 0 < p < 1 \)时,\( f \)为单调递减函数且\( \forall x \geq 0, f(x) \geq 0 \)。由\( 0 < p < 1 \)以及(1),可得\( \forall N > 0, \int_{[0, N]} f \)的值单调递增且无上界,故\( \sup_{N > 0} \int_{[0, N]} f = +\infty \)非有限,进而根据定理11.6.4,可得 \( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)发散。

当\( p = 1 \)时,\( f \)为单调递减函数且\( \forall x \geq 0, f(x) = \dfrac{1}{x + 1} \geq 0 \)。( 注:这里很遗憾,似乎要用到未定义的对数函数了 )定义函数\( F: [0, +\infty), \forall x \geq 0 \),令\( F(x) := \ln(x + 1) \),可得\( \forall x \geq 0, F'(x) = \dfrac{1}{x + 1} = f(x) \),于是根据定理11.9.4,可得\( \forall N > 0, \int_{[0, N]} f = F(N) - F(0) = \ln(N + 1) - \ln(0 + 1) = \ln(N + 1) - 0 = \ln(N + 1) \),注意到\( \forall N > 0, \int_{[0, N]} f \)的值无上界,故\( \sup_{N > 0} \int_{[0, N]} f = +\infty \)非有限,进而根据定理11.6.4,可得\( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)发散。

当\( p < 0 \)时,\( f \)单调递增且无上界,因此\( \lim_{n \to \infty} f(n) \neq 0 \),进而\( \sum_{n = 0}^{\infty} \dfrac{1}{(n + 1)^p} \)发散。

证毕。

章节11.8

勘误

作者对本章节的内容进行了很多勘误,改动比较大,有必要提一下,下面大部分内容都是从作者博客中复制过来的。

勘误1:

In the paragraph before Definition 11.8.1, remove the sentences after “defined as follows”. In Definition 11.8.1, add the hypothesis that \( \alpha \) be monotone increasing, and \( X \) be an interval that is closed in the sense of Definition 9.1.15 (之前\( X \)可以是任意包含\( I \)的定义域,现在只能是封闭区间了(即要是封闭集,还要是区间)), and alter the definition of \( \alpha[I] \) as follows.

  1. If \( I \) is empty, set \( \alpha[I]:=0 \).
  2. If \( I = \{ a \} \) is a point, set \( \alpha[\{ a \}] := (\lim_{x \to a^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \), with the convention that \( \lim_{x \to a^+; x \in X} \alpha(x) \) is \( \alpha(a) \) when \( a \) is the right endpoint of \( X \), and \( \lim_{x \to a^-; x \in X} \alpha(x) \) is \( \alpha(a) \) when \( a \) is the left endpoint of \( X \) (注意:前面说了\( X \)现在得是区间了,故有端点的概念了).
  3. If \( I = (a, b) \), set \( \alpha[(a, b)] := (\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \) (注意下,这里减的顺序和\( I = \{ a \} \)是反过来的, \( I = \{ a \} \)定义成那种比较反直觉的顺序是为了相消).
  4. If \( I = [a, b), (a, b] \), or \( [a, b] \), set \( \alpha[I] \) equal to \( \alpha(\{ a \}) + \alpha((a, b)), \alpha((a, b)) + \alpha(\{ b \}) \), or \( \alpha(\{ a \}) + \alpha((a, b)) + \alpha(\{ b \}) \) respectively.

After the definition, note that in the special case when \( \alpha \) is continuous(在\( X \)上的连续函数), the definition of \( \alpha[I] \) for \( I = (a, b), [a, b), (a, b], [a, b] \) simplifies to \( \alpha[I] = \alpha(b) - \alpha(a) \)(因为连续函数收敛,收敛则左右极限相等), and in this case one can extend the definition to functions \( \alpha \) that are continuous but not necessarily monotone increasing.

为什么还要求\( X \)是闭合集?根据作者博客中yalikes和作者等人的讨论,考虑到\( I := (0, 1), X := (0, 1), \alpha(x) = \dfrac{1}{1 - x} \),此时\( I \)是包含\( X \)的区间,但不是闭合集,而根据前面的定义,有\( \alpha[I] := (\lim_{x \to 1^-; x \in X} \alpha(x)) - (\lim_{x \to 0^+; x \in X} \alpha(x)) \),但\( \lim_{x \to 1^-; x \in X} \alpha(x) \)未定义(或者说\( = +\infty \),非实数,不能进行减法),故此时上面的定义是有问题的,因此要求\( X \)不仅要是区间,还要是闭合集,此时根据定义9.1.15以及\( I \subseteq X \),可得 \( X \)必然包含\( I \)的左右端点\( a, b \)(虽然\( I \)本身不一定包含它的端点,另外注意下,\( X \)不一定是有限区间,它还可能是\( [a, +\infty), (-\infty, +\infty) \)之类的区间),而\( \alpha \)是单调递增函数或者连续函数,如果\( \alpha \)是连续函数,则\( \lim_{x \to b^-; x \in X} \alpha(x), \lim_{x \to a^+; x \in X} \alpha(x) \)都有定义(且都是实数),如果\( \alpha \)是单调递增函数,则\( \alpha(a) \leq \lim_{x \to b^-; x \in X} \alpha(x) \leq \alpha(b) \) 是实数,同理\( \lim_{x \to a^+; x \in X} \alpha(x) \)也是实数。

勘误2:

In Example 11.8.2, restrict the domain of \( \alpha \) to \( [0, +\infty) \), and delete the example of \( \alpha[(-3,-2)] \).

勘误3:

In Example 11.8.6, restrict the domain of \( \alpha \) to \( [0, +\infty) \). In Lemma 11.8.4 and Definition 11.8.5, add the condition that \( X \) be an interval that is closed, and \( \alpha \) be monotone increasing or continuous.

勘误4:

After Example 11.8.7, delete the sentence “Up until now, our function… could have been arbitrary.”, and replace “defined on a domain” with “defined on an interval that is closed” (two occurrences).

勘误5:

The hint in Exercise 11.8.5 is no longer needed in view of other corrections and may be deleted.

练习11.8.1

题目:

Prove Lemma 11.8.4. (Hint: modify the proof of Theorem 11.1.13.)

Lemma 11.8.4的内容:

Let \( I \) be a bounded interval, let \( \alpha: X \to \mathbf{R} \) be a monotone increasing or continuous function defined on some closed interval \( X \) which contains \( I \), and let \( \mathbf{P} \) be a partition of \( I \). Then we have \( \alpha[I] = \sum_{J \in \mathbf{P}} \alpha[J] \).

注:

原本要讨论各种区间形式已经够琐碎了,现在勘误后单点和非单点的\( \alpha \)-长度形式还不一样,要讨论的更多了。

证明:

令\( n := \#(\mathbf{P}) \),对\( n \)进行数学归纳。

当\( n = 0 \)时,\( \mathbf{P} = \empty, I = \empty \),因此\(\alpha[I] = 0 = \sum_{J \in \mathbf{P}} \alpha[J] \)成立。

归纳假设当\( n = k \)时成立,当\( n = k + 1 \)时:

如果\( I = \empty \),则\( \mathbf{P} = \empty \)或\( \mathbf{P} = \{ \empty \} \),两种情况均有\(\alpha[I] = 0 = \sum_{J \in \mathbf{P}} \alpha[J] \)。

如果\( I \)仅包含单个点\( = \{ a \} \),则\( \mathbf{P} = \{ \{ a \} \} \)或\( \mathbf{P} = \{ \{ a \}, \empty \} \),两种情况均有\(\alpha[I] = \alpha[\{ a \}] = \sum_{J \in \mathbf{P}} \alpha[J] \)。

接下来我们讨论\( I \)不为空集且不仅包含单个点的情况,此时\( I \)会为\( [a, b], (a, b], (a, b), [a, b) \)形式的其中一种且\( a < b \)。

如果\( b \in I \),则\( I \)会为\( [a, b], (a, b] \)形式的其中一种且\( a < b \),由\( \mathbf{P} \)是\( I \)的区间,可得\( \exists K \in \mathbf{P}, b \in K \), \( K \)会为\( (c, b], [c, b] \)形式中的其中一种且\( a \leq c \leq b \) (注:特别的,当\( c = b \)时,\( K = [c, c] = \{ c \} \)),进而可得 \( I - K \)会为\( [a, c], (a, c], (a, c), [a, c) \)形式中的其中一种。由于\( \mathbf{P} \)是\( I \)的区间,因此\( \mathbf{P} - \{ K \} \)是\( I - K \)的区间且\( \#(\mathbf{P} - \{ K \}) = n \),根据归纳假设,有\( \alpha[I - K] = \sum_{J \in \mathbf{P} - \{ K \}} \alpha[J] \) (1) 。我们考虑\( K = [c, b] \)形式的情况,另外一种情况类似:

  1. 如果\( c = b \),则\( K = \{ c \} \)且\( I - K = [a, c) \)或\( I - K = (a, c) \),此时\( \alpha[K]= (\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x)) \)。针对\( I - K \)进行分类讨论:
    1. 如果\( I - K = [a, c) \),则\( \alpha[I - K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to a^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) = (\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),于是有\( \alpha[I - K] + \alpha[K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) + ((\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),而\( \alpha[I] = \alpha[ [a, b] ] = ((\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to b^-; x \in X} \alpha(x))) + ((\lim_{x \to a^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),可得\( \alpha[I] = \alpha[I - K] + \alpha[K] \)。
    2. 如果\( I - K = (a, c) \),则\( \alpha[I - K] = (\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),于是有\( \alpha[I - K] + \alpha[K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = (\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),而\( \alpha[I] = \alpha[ (a, b] ] = ((\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to b^-; x \in X} \alpha(x))) = (\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),可得\( \alpha[I] = \alpha[I - K] + \alpha[K] \)。
  2. 如果\( c \neq b \),则\( K = [c, b] \)(前面说了,\( K = (c, b] \)的情况类似,这里不讨论),且\( I - K = [a, c) \)或\( I - K = (a, c) \),此时\( \alpha[K]= ((\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to c^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to b^-; x \in X} \alpha(x))) + ((\lim_{x \to c^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x)) \)。针对\( I - K \)进行分类讨论:
    1. 如果\( I - K = [a, c) \),则\( \alpha[I - K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to a^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) = (\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),于是有\( \alpha[I - K] + \alpha[K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),而\( \alpha[I] = \alpha[ [a, b] ] = ((\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to b^-; x \in X} \alpha(x))) + ((\lim_{x \to a^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^-; x \in X} \alpha(x)) \),可得\( \alpha[I] = \alpha[I - K] + \alpha[K] \)。
    2. 如果\( I - K = (a, c) \),则\( \alpha[I - K] = (\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),于是有\( \alpha[I - K] + \alpha[K] = ((\lim_{x \to c^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to c^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),而\( \alpha[I] = \alpha[ (a, b] ] = ((\lim_{x \to b^-; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x))) + ((\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to b^-; x \in X} \alpha(x))) = (\lim_{x \to b^+; x \in X} \alpha(x)) - (\lim_{x \to a^+; x \in X} \alpha(x)) \),可得\( \alpha[I] = \alpha[I - K] + \alpha[K] \)。

综上,有\( \alpha[I] = \alpha[I - K] + \alpha[K] \),加上(1),可得\( \alpha[I] = \sum_{J \in \mathbf{P}} \alpha[J] \)。

如果\( b \notin I \),则\( I \)会为\( (a, b), [a, b) \)形式的其中一种且\( a < b \),此时根据练习11.1.3,可得\( \exists K \in \mathbf{P}, K \)的形式为\( (c, b) \)或\( [c, b) \) 中的其中一种且\( a \leq c \leq b \),这意味着\( I - K \)会为 \( [a, c], (a, c], (a, c), [a, c) \)形式中的其中一种,剩余的部分跟上面同理,易证 \( \alpha[I] = \sum_{J \in \mathbf{P}} \alpha[J] \)成立。

至此可得,当\( n = k + 1 \)时也成立,归纳完毕。

证毕。

练习11.8.2

题目:

State and prove a version of Proposition 11.2.13 for the Riemann-Stieltjes integral.

Proposition 11.2.13的黎曼-斯蒂尔杰斯积分版本:

(Piecewise constant Riemann-Stieltjes integral is independent of partition). Let \( I \) be a bounded interval, let \( \alpha: X \to \mathbf{R} \) be a monotone increasing or continuous function defined on some closed interval \( X \) which contains \( I \), and let \( f: I \to \mathbf{R} \) be a function. Suppose that \( \mathbf{P} \) and \( \mathbf{P'} \) are partitions of \( I \) such that \( f \) is piecewise constant both with respect to \( \mathbf{P} \) and with respect to \( \mathbf{P'} \). Then \( p.c. \int_{[\mathbf{P}]} f \mathop{} d \alpha = p.c. \int_{[\mathbf{P'}]} f \mathop{} d \alpha \).

注1:

和Proposition 11.2.13的证明基本一样,直接搬过来改下。

注2:

给定一个区间\( J \)(或者其他区间符号),如果\( f \)在\( J \)上为常数函数,则我们约定\( c_J \)为\( f \)在\( J \)上的常数值,不再显式引入符号\( c_J \)。

证明:

因为\( \mathbf{P}, \mathbf{P'} \)均是\( I \)的分区,根据引理11.1.18,可得\( \mathbf{P} \# \mathbf{P'} \)也是\( I \)的分区且比\( \mathbf{P}, \mathbf{P'} \)都精细。

\( \forall J \in \mathbf{P} \),令\( \mathbf{P_J} := \{ J \cap K : K \in \mathbf{P'} \} \),我们证明\( \mathbf{P_J} \)为\( J \)的分区:首先证明\( \bigcup_{L \in \mathbf{P_J}} L = J \),分情况讨论:

  1. 如果\( J = \emptyset \),则\( \bigcup_{L \in \mathbf{P_J}} L = \emptyset = J \)。
  2. 如果\( J \neq \emptyset \),则根据\( \mathbf{P_J} \)的定义,明显有\( \bigcup_{L \in \mathbf{P_J}} L \subseteq J \),再证明下\( J \subseteq \bigcup_{L \in \mathbf{P_J}} L \)就行,因为\( \mathbf{P} \# \mathbf{P'} \)比\( \mathbf{P} \)精细,因此\( \exists J_L \in \mathbf{P}, K_L \in \mathbf{P'}, J \subseteq J_L \cap K_L \),又\( \mathbf{P} \)中的区间互不相交,因此\( J_L = J \)(注意,\( J \neq \emptyset \)),于是有\( J \subseteq J \cap K_L \),加上\( J \cap K_L \subseteq \bigcup_{L \in \mathbf{P_J}} L \),可得\( J \subseteq \bigcup_{L \in \mathbf{P_J}} L \),综上,有\( \bigcup_{L \in \mathbf{P_J}} L = J \)。

接着证明\( \mathbf{P_J} \)中的区间互不相交,假设\( \exists L_1, L_2 \in \mathbf{P_J}, L_1 \cap L_2 \neq \emptyset \),则由于\( \mathbf{P_J} \subseteq \mathbf{P} \# \mathbf{P'} \),可得\( L_1, L_2 \in \mathbf{P} \# \mathbf{P'} \),而\( L_1 \cap L_2 \neq \emptyset \),这与\( \mathbf{P} \# \mathbf{P'} \)是\( I \)的分区矛盾, 因此假设不成立,有\( \mathbf{P_J} \)中的区间互不相交。

综上,有\( \mathbf{P_J} \)为\( J \)的分区。

根据引理7.1.13,可得\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha = \sum_{L \in \mathbf{P} \# \mathbf{P'}} c_L \alpha[L] = \sum_{J \in \mathbf{P}} (\sum_{K \in \mathbf{P'}} c_{J \cap K} \alpha[J \cap K]) \) (1)

\( \forall J \in \mathbf{P} \),因为\( f \)针对\( \mathbf{P} \)为分段常数函数,因此\( f \)在\( J \)上为常数函数(注意,没有“分段”两个字) (2) ,针对\( \forall K \in \mathbf{P'} \),由\( J \cap K \subseteq J \)以及(2),可得\( c_{J \cap K} = c_J \),于是有\( \sum_{K \in \mathbf{P'}} c_{J \cap K} \alpha[J \cap K] = \sum_{K \in \mathbf{P'}} c_J \alpha[J \cap K] \),而\( \sum_{K \in \mathbf{P'}} c_J \alpha[J \cap K] = \sum_{L \in \mathbf{P_J}} c_J \alpha[L] \),因此可得\( \sum_{K \in \mathbf{P'}} c_{J \cap K} \alpha[J \cap K] = \sum_{L \in \mathbf{P_J}} c_J \alpha[L] = c_J \sum_{L \in \mathbf{P_J}} \alpha[L] \),再根据引理11.8.4以及\( \mathbf{P_J} \)为\( J \)的分区,可得\( \sum_{L \in \mathbf{P_J}} \alpha[L] = \alpha[J] \),于是有\( \sum_{K \in \mathbf{P'}} c_{J \cap K} \alpha[J \cap K] = c_J \sum_{L \in \mathbf{P_J}} \alpha[L] = c_J \alpha[J] \) (3)

根据(1)和(3),\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha = \sum_{J \in \mathbf{P}} (\sum_{K \in \mathbf{P'}} c_{J \cap K} \alpha[J \cap K]) = \sum_{J \in \mathbf{P}} c_J \alpha[J] = p.c. \int_{[\mathbf{P}]} f \mathop{} d \alpha \)。

同理易证\( p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha = p.c. \int_{[\mathbf{P'}]} f \mathop{} d \alpha \),于是可得\( p.c. \int_{[\mathbf{P}]} f \mathop{} d \alpha = p.c. \int_{[\mathbf{P'}]} f \mathop{} d \alpha \)。

证毕。

练习11.8.3

题目:

State and prove a version of Theorem 11.2.16 for the Riemann-Stieltjes integral.

Theorem 11.2.16的黎曼-斯蒂尔杰斯积分版本:

(Laws of integration). Let \( I \) be a bounded interval, let \( \alpha: X \to \mathbf{R} \) be a monotone increasing function defined on some closed interval \( X \) which contains \( I \), and let \( f: I \to \mathbf{R} \) and \( g: I \to \mathbf{R} \) be piecewise constant functions on \( I \).

  1. We have \( p.c. \int_{I} (f + g) \mathop{} d \alpha = p.c. \int_{I} f \mathop{} d \alpha + p.c. \int_{I} g \mathop{} d \alpha \).
  2. For any real number \( c \), we have \( p.c. \int_{I} (cf) \mathop{} d \alpha = c (p.c. \int_{I} f) \mathop{} d \alpha \).
  3. We have \( p.c. \int_{I} (f - g) \mathop{} d \alpha = p.c. \int_{I} f \mathop{} d \alpha - p.c. \int_{I} g \mathop{} d \alpha \).
  4. If \( f(x) \geq 0 \) for all \( x \in I \), then \( p.c. \int_{I} f \mathop{} d \alpha \geq 0 \).
  5. If \( f(x) \geq g(x) \) for all \( x \in I \), then \( p.c. \int_{I} f \mathop{} d \alpha \geq p.c. \int_{I} g \mathop{} d \alpha \).
  6. If \( f \) is the constant function \( f(x) = c \) for all \( x \in I \), then \( p.c. \int_{I} f \mathop{} d \alpha = c \alpha[I] \).
  7. Let \( J \) be a bounded interval containing \( I \) (i.e. \( I \subseteq J \)), and let \( F: J \to \mathbf{R} \) be the function \( F(x) := \begin{cases} f(x) & \text{ if } x \in I \\ 0 & \text{ if } x \notin I \end{cases} \), Then \( F \) is piecewise constant on \( J \), and \( p.c. \int_{J} F \mathop{} d \alpha = p.c. \int_{I} f \mathop{} d \alpha \).
  8. Suppose that \( \{ J, K \} \) is a partition of \( I \) into two intervals \( J \) and \( K \). Then the functions \( f|_J: J \to \mathbf{R} \) and \( f|_K: K \to \mathbf{R} \) are piecewise constant on \( J \) and \( K \) respectively, and we have \( p.c. \int_{I} f \mathop{} d \alpha = p.c. \int_{J} f|_J \mathop{} d \alpha + p.c. \int_{K} f|_K \mathop{} d \alpha \).

注:

和Proposition 11.2.16的证明基本一样,直接搬过来改下。

证明:

因为\( f, g \)均为\( I \)上的分段常数函数,因此\( \exists I \)的分区\( \mathbf{P}, \mathbf{P'} \)使得 \( f \)针对\( \mathbf{P} \)为分段常数函数, \( g \)针对\( \mathbf{P'} \)为分段常数函数,根据引理11.1.18和引理11.2.7,可得\( f, g \)均针对\( \mathbf{P} \# \mathbf{P'} \)为分段常数函数, \( \forall K \in \mathbf{P} \# \mathbf{P'} \),令\( c_{K_f} \)为\( f \)在\( K \)上的常数值,令\( c_{K_g} \)为\( g \)在\( K \)上的常数值。

证明1:

\( p.c. \int_{I} (f + g) \mathop{} d \alpha = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} (f + g) \mathop{} d \alpha = \sum_{K \in \mathbf{P} \# \mathbf{P'}} (c_{K_f} + c_{K_g}) \alpha[K] = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f} \alpha[K]) + (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_g} \alpha[K]) = (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha) + (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} g \mathop{} d \alpha) = (p.c. \int_{I} f \mathop{} d \alpha) + (p.c. \int_{I} g \mathop{} d \alpha) \)。

证明2:

\( \forall c \in \mathbf{R}, p.c. \int_{I} (cf) \mathop{} d \alpha = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} (cf) \mathop{} d \alpha = \sum_{K \in \mathbf{P} \# \mathbf{P'}} cc_{K_f} \alpha[K] = c (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f} \alpha[K]) = c (p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha) = c (p.c. \int_{I} f \mathop{} d \alpha) \)。

证明3:

根据2,有\( p.c. \int_{I} (-g) \mathop{} d \alpha = -(p.c. \int_{I} g \mathop{} d \alpha) \),再根据1,可得\( p.c. \int_{I} (f - g) \mathop{} d \alpha = p.c. \int_{I} (f + (-g)) \mathop{} d \alpha = (p.c. \int_{I} f \mathop{} d \alpha) + (p.c. \int_{I} (-g) \mathop{} d \alpha) = (p.c. \int_{I} f \mathop{} d \alpha) - (p.c. \int_{I} g \mathop{} d \alpha) \)。

证明4:

如果\( \forall x \in I, f(x) \geq 0 \),则\( \forall K \in \mathbf{P} \# \mathbf{P'}, c_{K_f} \geq 0 \),又\( \alpha \)单调递增,因此\( \alpha[K] \geq 0 \),此时有\( p.c. \int_{I} f \mathop{} d \alpha = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f} \alpha[K] \geq 0 \)。

证明5:

如果\( \forall x \in I, f(x) \geq g(x) \),则\( \forall x \in I, (f - g)(x) = f(x) - g(x) \geq 0 \),此时根据4以及3,可得\( p.c. \int_{I} (f - g) \mathop{} d \alpha = (p.c. \int_{I} f \mathop{} d \alpha) - (p.c. \int_{I} g \mathop{} d \alpha) \geq 0 \),进而可得\( p.c. \int_{I} f \mathop{} d \alpha \geq p.c. \int_{I} g \mathop{} d \alpha \)。

证明6:

由\( \mathbf{P} \# \mathbf{P'} \)是\( I \)的分区以及引理11.8.4,可得 \( \sum_{K \in \mathbf{P} \# \mathbf{P'}} \alpha[K] = \alpha[I] \)。

如果\( \exists c \in \mathbf{R}, \forall x \in I, f(x) = c \),则\( \forall K \in \mathbf{P} \# \mathbf{P'}, c_{K_f} = c \),此时\( p.c. \int_{I} f \mathop{} d \alpha = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f} \alpha[K] = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c \alpha[K] = c (\sum_{K \in \mathbf{P} \# \mathbf{P'}} \alpha[K]) = c \alpha[I] \)。

证明7:

如果\( I = \emptyset \),则\( \forall x \in J, F(x) = 0 \),此时易得\( p.c. \int_{J} F \mathop{} d \alpha = p.c. \int_{I} f \mathop{} d \alpha = 0 \)。

下面考虑\( I \neq \emptyset \)的情况:

因为\( I \)为有界区间以及\( I \neq \emptyset \),因此\( \exists a_I \leq b_I \in \mathbf{R} \) 使得\( I \)为\( [a_I, b_I], (a_I, b_I), [a_I, b_I), (a_I, b_I] \)中的其中一种。因为\( I \subseteq J \),因此\( J \neq \emptyset \),又因为\( J \)为有界区间,因此\( \exists a_J \leq b_J \in \mathbf{R} \) 使得\( J \)为\( [a_J, b_J], (a_J, b_J), [a_J, b_J), (a_J, b_J] \)中的其中一种。因为\( I \subseteq J \),因此\( a_J \leq a_I, b_I \leq b_J \) (考虑下,如果\( a_J > a_I \)或\( b_I > b_J \),则不管两个区间是哪种形式,均会违反\( I \subseteq J \)),可得\( J - I = J_l \cup J_r \),其中\( J_l, J_r \)均为有界区间,\( J_l \cap J_r = \emptyset \),并且,\( J_l \)为\( [a_J, a_I], (a_J, a_I), [a_J, a_I), (a_J, a_I] \)中的其中一种, \( J_r \)为\( [b_I, b_J], (b_I, b_J), [b_I, b_J), (b_I, b_J] \)中的其中一种,令\( \mathbf{P_J} := (\mathbf{P} \# \mathbf{P'}) \cup \{ J_l, J_r \} \),易得\( \mathbf{P_J} \)为\( J \)的分区且\( F \)针对\( \mathbf{P_J} \)为分段常数函数。 \( \forall K \in \mathbf{P_J} \),令\( c_{K_F} \)为\( F \)在\( K \)上的常数值。 \( \forall x \in J_l \cup J_r = J - I \),有\( x \notin I \),因此\( F(x) = 0 \),进而可得\( F \)在\( J_l, J_r \)上的常数值\( c_{J_l}, c_{J_r} \)均\( = 0 \)。 \( \forall x \in \bigcup_{K \in \mathbf{P} \# \mathbf{P'}} K \),有\( x \in I \),因此\( F(x) = f(x) \),进而可得\( \forall K \in \mathbf{P} \# \mathbf{P'} \), \( F \)在\( K \)上的常数值\( c_{K_F} = \)\( f \)在\( K \)上的常数值\( c_{K_f} \),此时可得\( p.c. \int_{J} F \mathop{} d \alpha = p.c. \int_{[\mathbf{P_J}]} F \mathop{} d \alpha = \sum_{K \in \mathbf{P_J}} c_{K_F} \alpha[K] = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F} \alpha[K]) + (c_{J_l} \alpha[J_l]) + (c_{J_r} \alpha[J_r]) = (\sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F} \alpha[K]) + (0 \alpha[J_l]) + (0 \alpha[J_r]) = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_F} \alpha[K] = \sum_{K \in \mathbf{P} \# \mathbf{P'}} c_{K_f} \alpha[K] = p.c. \int_{[\mathbf{P} \# \mathbf{P'}]} f \mathop{} d \alpha = p.c. \int_{I} f \mathop{} d \alpha \)。

证明8:

注:这里为了方便,我们shadow之前定义的\( \mathbf{P}, \mathbf{P'} \)符号。

因为\( f \)为\( I \)上的分段常数函数,因此\( \exists I \)的分区\( \mathbf{P} \)使得 \( f \)针对\( \mathbf{P} \)为分段常数函数。又\( \{ J, K \} \)也是\( I \)的分区,因此根据引理11.1.18,可得\( \mathbf{P'} := \mathbf{P} \# \{ J, K \} \)也是\( I \)的分区且比\( \mathbf{P}, \{ J, K \} \)都精细。再根据引理11.2.7,可得\( f \)针对\( \mathbf{P'} \)为分段常数函数。

令\( \mathbf{P_J} := \{ L \cap J : L \in \mathbf{P} \} \),我们证明下\( \mathbf{P_J} \)是\( J \)的分区: \( \forall x \in J \),有\( x \in I \),又\( \mathbf{P} \)是\( I \)的分区,因此\( \exists L_x \in \mathbf{P}, x \in L_x \),加上\( x \in J \),可得\( x \in L_x \cap J \),进而\( x \in \mathbf{P_J} \)。接着证明\( \mathbf{P_J} \)中的区间互不相交,假设\( \mathbf{P_J} \)中存在相交的的区间,即\( \exists L_1, L_2 \in \mathbf{P}, (L_1 \cap J) \cap (L_2 \cap J) \neq \emptyset \),这意味着\( \exists x \)满足\( x \in L_1 \)且\( x \in L_2 \),这和\( \mathbf{P} \)为\( I \)的区间矛盾,因此假设不成立,有\( \mathbf{P_J} \)中的区间互不相交。至此,我们证明了\( \mathbf{P_J} \)是\( J \)的分区。同理可证\( \mathbf{P_K} := \{ L \cap K : L \in \mathbf{P} \} \),是\( K \)的分区。

由\( \mathbf{P_J} \subseteq \mathbf{P'} \)以及\( f \)针对\( \mathbf{P'} \)为分段常数函数,可得\( f \)在\( \mathbf{P_J} \)的区间上的函数值均为常数值,而\( f_J \)在这些区间上的函数值和\( f \)相同,因此\( f_J \)也在\( \mathbf{P_J} \)的区间上的函数值均为常数值,即\( f_J \)为\( \mathbf{P_J} \)上的分段常数函数。同理可证\( f_K \)为\( \mathbf{P_K} \)上的分段常数函数。

注意到\( \mathbf{P'} = \mathbf{P_J} \cup \mathbf{P_K} \)且 \( \mathbf{P_J} \cap \mathbf{P_K} = \emptyset \),可得 \( p.c. \int_{I} f \mathop{} d \alpha = p.c. \int_{[\mathbf{P'}]} f \mathop{} d \alpha = \sum_{L \in \mathbf{P'}} c_L \alpha[L] = (\sum_{L_J \in \mathbf{P_J}} c_{L_J} \alpha[L_J]) + (\sum_{L_K \in \mathbf{P_K}} c_{L_K} \alpha[L_K]) = (p.c. \int_{J} f|_J \mathop{} d \alpha) + (p.c. \int_{K} f|_K \mathop{} d \alpha) \)。

证毕。

额外定义11.8.1、额外引理11.8.2、额外引理11.8.3

注:

下面的额外定义、额外引理是为了证明练习11.8.4准备的。在这几个额外引理中,我们限制\( \alpha \)为单调递增函数,不然几个额外引理似乎不成立,与之相对应的,练习11.8.4的\( \alpha \)也会限制为单调递增函数。额外定义11.8.1 似乎可以不限制\( \alpha \)为单调递增函数(即\( \alpha \)为单调递增函数或者连续函数),但为了一致性,这里统一进行限制。

额外定义11.8.1(Riemann-Stieltjes sums):

Let \( f: I \to \mathbf{R} \) be a bounded function on a bounded interval \( I \), let \( \alpha: X \to \mathbf{R} \) be a monotone increasing function defined on some closed interval \( X \) which contains \( I \), and let \( \mathbf{P} \) be a partition of \( I \). We define the upper Riemann-Stieltjes sum \( U(f, \mathbf{P}, \alpha) \) and the lower Riemann-Stieltjes sum \( L(f, \mathbf{P}, \alpha) \) by \( U(f, \mathbf{P}, \alpha) := \sum_{J \in \mathbf{P} : J \neq \empty} (\sup_{x \in J} f(x)) \alpha[J] \) and \( L(f, \mathbf{P}, \alpha) := \sum_{J \in \mathbf{P} : J \neq \empty} (\inf_{x \in J} f(x)) \alpha[J] \).

额外引理11.8.2:

Let \( f: I \to \mathbf{R} \) be a bounded function on a bounded interval \( I \), let \( g \) be a function which majorizes \( f \) and which is piecewise constant with respect to some partition \( \mathbf{P} \) of \( I \), let \( \alpha: X \to \mathbf{R} \) be a monotone increasing function defined on some closed interval \( X \) which contains \( I \). Then \( p.c. \int_{I} g \mathop{} d \alpha \geq U(f, \mathbf{P}, \alpha) \). Similarly, if \( h \) is a function which minorizes \( f \) and is piecewise constant with respect to \( \mathbf{P} \), then \( p.c. \int_{I} h \mathop{} d \alpha \leq L(f, \mathbf{P}, \alpha) \).

证明:

证明如果分段常数函数\( g \)主要于\( f \),则\( p.c. \int_{I} g \mathop{} d \alpha \geq U(f, \mathbf{P}, \alpha) \):

如果\( I = \emptyset \),则\( p.c. \int_{I} g \mathop{} d \alpha = 0 \geq 0 = U(f, \mathbf{P}, \alpha) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall J \neq \emptyset \in \mathbf{P} \),因为\( g \)主要于\( f \),因此\( \forall x \in J, g(x) \geq f(x) \),进而\( \sup_{x \in J} g(x) \geq \sup_{x \in J} f(x) \),而\( \forall x \in J, g(x) \)均\( = g \)在\( J \)上的常数值\( c_J \),因此\( c_J = \sup_{x \in J} g(x) \geq \sup_{x \in J} f(x) \),又\( \alpha \)为单调递增函数,可得\( \alpha[I] \geq 0 \),进而可得\( c_J \alpha[J] \geq (\sup_{x \in J} f(x)) \alpha[J] \)。综上可得,\( p.c. \int_{I} g \mathop{} d \alpha = \sum_{J \in \mathbf{P}} c_J \alpha[J] = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J \alpha[J]) + (\sum_{J \in \mathbf{P} : J = \emptyset} c_J \alpha[J]) = (\sum_{J \in \mathbf{P} : J \neq \emptyset} c_J \alpha[J]) + 0 = \sum_{J \in \mathbf{P} : J \neq \emptyset} c_J \alpha[J] \geq \sum_{J \in \mathbf{P} : J \neq \emptyset} (\sup_{x \in J} f(x)) \alpha[J] \),即\( p.c. \int_{I} g \mathop{} d \alpha \geq U(f, \mathbf{P}, \alpha) \)。

同理易证如果分段常数函数\( h \)次要于\( f \),则\( p.c. \int_{I} h \mathop{} d \alpha \leq L(f, \mathbf{P}, \alpha) \)。

证毕。

额外引理11.8.3:

Let \( f: I \to \mathbf{R} \) be a bounded function on a bounded interval \( I \), let \( \alpha: X \to \mathbf{R} \) be a monotone increasing function defined on some closed interval \( X \) which contains \( I \). Then \( \overline{\int}_{I} f \mathop{} d \alpha = \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{ is a partition of } I \}) \) and \( \underline{\int}_{I} f \mathop{} d \alpha = \sup(\{ L(f, \mathbf{P}, \alpha) : \mathbf{P} \text{ is a partition of } I \}) \).

证明:

证明\( \overline{\int}_{I} f \mathop{} d \alpha = \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \):

如果\( I = \emptyset \),则\( \overline{\int}_{I} f \mathop{} d \alpha = 0 = \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

下面考虑\( I \neq \emptyset \)的情况:

\( \forall g \in \{ p.c. \int_{I} g \mathop{} d \alpha : g \text{为} I \text{上主要于} f \text{的分段常数函数} \} \),有\( g \)主要于\( f \),\( \forall I \)的分区\( \mathbf{P} \),根据额外引理11.8.2,可得\( p.c. \int_{I} g \mathop{} d \alpha \geq U(f, \mathbf{P}, \alpha) \),于是可得\( p.c. \int_{I} g \mathop{} d \alpha \)为 \( \{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \} \)的上界,进而可得\( p.c. \int_{I} g \mathop{} d \alpha \geq \sup(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \geq \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \) (重点在于\( p.c. \int_{I} g \mathop{} d \alpha \geq \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \))。综上,可得\( \overline{\int}_{I} f \mathop{} d \alpha = \inf(\{ p.c. \int_{I} g \mathop{} d \alpha : g \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \geq \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

还差证明\( \overline{\int}_{I} f \mathop{} d \alpha \leq \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \),假设\( \overline{\int}_{I} f \mathop{} d \alpha > \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \),则\( \exists I \)的分区\( \mathbf{P'} \),使得\( U(f, \mathbf{P'}, \alpha) < \overline{\int}_{I} f \mathop{} d \alpha \) (1) ,定义函数\( h: I \to \mathbf{R}, \forall y \in I \),因为\( \mathbf{P'} \)为\( I \)的分区,因此\( \exists \)唯一的\( J_y \in \mathbf{P'}, y \in J_y \)(注意,这意味着\( J_y \)不为空),令\( h(y) := \sup_{x \in J_y} f(x) \),易得\( h \)为\( I \)上的分段常数函数(属于同一个区间\( J \)内的点的函数值均为\( \sup_{x \in J} f(x) \))且\( p.c. \int_{I} h \mathop{} d \alpha = p.c. \int_{[\mathbf{P'}]} h \mathop{} d \alpha = \sum_{J \in \mathbf{P'} : J \neq \emptyset} (\sup_{x \in J} f(x)) \alpha[J] = U(f, \mathbf{P'}, \alpha) \) (2) ,又\( \forall x \in I, h(x) \geq f(x) \),因此也有\( h \)主要于\( f \),于是可得\( p.c. \int_{I} h \mathop{} d \alpha \in \{ p.c. \int_{I} g \mathop{} d \alpha : g \text{为} I \text{上主要于} f \text{的分段常数函数} \} \),进而可得\( \overline{\int}_{I} f \mathop{} d \alpha = \inf(\{ p.c. \int_{I} g \mathop{} d \alpha : g \text{为} I \text{上主要于} f \text{的分段常数函数} \}) \leq p.c. \int_{I} h \mathop{} d \alpha \),加上(2),可得\( \overline{\int}_{I} f \mathop{} d \alpha \leq U(f, \mathbf{P'}, \alpha) \),这和(1)矛盾,因此假设不成立,有\( \overline{\int}_{I} f \mathop{} d \alpha \leq \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

综上,可得\( \overline{\int}_{I} f \mathop{} d \alpha = \inf(\{ U(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

同理易证\( \underline{\int}_{I} f \mathop{} d \alpha = \sup(\{ L(f, \mathbf{P}, \alpha) : \mathbf{P} \text{为} I \text{的分区} \}) \)。

证毕。

练习11.8.4

题目:

State and prove a version of Theorem 11.5.1 for the Riemann-Stieltjes integral. (Hint: one has to be careful with the proof; the problem here is that some of the references to the length of \( |J_k| \) should remain unchanged, and other references to the length of \( |J_k| \) should be changed to the \( \alpha \)-length \( \alpha[Jk] \) - basically, all of the occurrences of \( |J_k| \) which appear inside a summation should be replaced with \( \alpha[J_k] \), but the rest should be unchanged.)

Theorem 11.5.1的黎曼-斯蒂尔杰斯积分版本:

Let \( I \) be a bounded interval, let \( \alpha: X \to \mathbf{R} \) be a monotone increasing function defined on some closed interval \( X \) which contains \( I \), and let \( f \) be a function which is uniformly continuous on \( I \). Then \( f \) is Riemann-Stieltjes integrable.

证明:

由定理9.9.15,我们知道\( f \)是有界的,现在我们需要证明\( \underline{\int}_{I} f \mathop{} d \alpha = \overline{\int}_{I} f \mathop{} d \alpha \)。

如果\( I \)为空集或者仅包含单个点,则\( f \)明显黎曼-斯蒂尔杰斯可积。

接下来我们讨论\( I \)不为空集且不仅包含单个点的情况,此时\( I \)会为\( [a, b], (a, b], (a, b), [a, b) \)形式的其中一种且\( a < b \)。

\( \forall \epsilon > 0 \),由\( f \)在\( I \)上一致连续,可得 \( \exists \delta > 0, \forall x, y \in I, |x - y| < \delta, |f(x) - f(y)| < \epsilon \),根据阿基米德原理,可得\( \exists N > 0 \in \mathbf{Z} \)满足 \( (b - a) / N < \delta \)。

易得我们可以将\( I \)分区成\( N \)个区间:\( J_1, \dots, J_N \),每个区间的长度为\( (b - a) / N \)(注:需要分\( [a, b], (a, b], (a, b), [a, b) \) 四种情况讨论,每种情况略微不同,但是差别不大),根据额外引理11.8.3,可得\( \overline{\int}_{I} f \mathop{} d \alpha \leq \sum_{k = 1}^{N} (\sup_{x \in J_k} f(x)) \alpha[J_k] \) 以及\( \underline{\int}_{I} f \mathop{} d \alpha \geq \sum_{k = 1}^{N} (\inf_{x \in J_k} f(x)) \alpha[J_k] \),进而可得\( \overline{\int}_{I} f \mathop{} d \alpha - \underline{\int}_{I} f \mathop{} d \alpha \leq \sum_{k = 1}^{N} (\sup_{x \in J_k} f(x) - \inf_{x \in J_k} f(x)) \alpha[J_k] \) (1)

由于\( |J_k| = \dfrac{b - a}{N} < \delta \),因此我们有\( \forall x, y \in J_k, |f(x) - f(y)| < \epsilon \),特别的,可得\( \forall x, y \in J_k, f(x) < f(y) + \epsilon \),针对\( x \)取上界,可得\( \forall y \in J_k, \sup_{x \in J_k} f(x) \leq f(y) + \epsilon \),继续针对\( y \)取下界,可得\( \sup_{x \in J_k} f(x) \leq \inf_{y \in J_k} f(y) + \epsilon \),该不等式结合(1),可得\( \overline{\int}_{I} f \mathop{} d \alpha - \underline{\int}_{I} f \mathop{} d \alpha \leq \sum_{k = 1}^{N} \epsilon \alpha[J_k] \),此时根据引理11.8.4,可得 \( \overline{\int}_{I} f \mathop{} d \alpha - \underline{\int}_{I} f \mathop{} d \alpha \leq \epsilon \alpha[I] \),加上\( \epsilon \)可以任意小,可得\( \overline{\int}_{I} f \mathop{} d \alpha = \underline{\int}_{I} f \mathop{} d \alpha \),即\( f \)黎曼-斯蒂尔杰斯可积。

证毕。

练习11.8.5

题目:

Let \( \text{sgn}: \mathbf{R} \to \mathbf{R} \) be the signum function \( \text{sgn}(x) := \begin{cases} 1 & \text{ when } x > 0 \\ 0 & \text{ when } x = 0 \\ -1 & \text{ when } x < 0 \end{cases} \). Let \( f: [-1, 1] \to \mathbf{R} \) be a continuous function. Show that \( f \) is Riemann-Stieltjes integrable with respect to \( \text{sgn} \), and that \( \int_{[-1, 1]} f \mathop{} d \text{sgn} = 2f(0) \). (Hint: for every \( \epsilon > 0 \), find piecewise constant functions majorizing and minorizing \( f \) whose Riemann-Stieltjes integral is \( \epsilon \)-close to \( 2f(0) \).)

证明:

根据引理9.6.3,可得\( f \)有界,为了证明\( f \)黎曼-斯蒂尔杰斯可积且 \( \int_{[-1, 1]} f \mathop{} d \text{sgn} = 2f(0) \),我们还得证明\( \underline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} = \overline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} = 2f(0) \)。

\( \forall \epsilon > 0 \):由\( f \)连续,可得 \( \exists \delta_0 > 0, \forall x \in [-1, 1], |x - 0| < \delta_0, |f(x) - f(0)| < \dfrac{\epsilon}{2} \),令\( \delta := \min(\delta_0, \dfrac{1}{2}) \),可得\( 0 < \delta \leq \dfrac{1}{2} \) 且\( \forall x \in [-1, 1], |x - 0| < \delta, |f(x) - f(0)| < \dfrac{\epsilon}{2} \) (1) 。构造分段常数函数\( g_{\epsilon}: [-1, 1] \to \mathbf{R}, \forall x \in [-1, 1] \),如果\( |x - 0| < \delta \),即\( -\delta < x < \delta \),则令\( g_{\epsilon} := \sup_{x \in (-\delta, \delta)} f(x) \),如果\( x \in [\delta, 1] \)(注:\( 0 < \delta \leq \dfrac{1}{2} \)),则令\( g_{\epsilon} := \sup_{x \in [\delta, 1]} f(x) \),如果\( x \in [-1, -\delta] \),则令\( g_{\epsilon} := \sup_{x \in [-1, -\delta]} f(x) \),明显\( g_{\epsilon} \)是主要于\( f \)的分段常数函数,而且\( p.c. \int_{[-1, 1]} g_{\epsilon} \mathop{} d \text{sgn} = (\sup_{x \in [-1, -\delta]} f(x))(\text{sgn}[ [-1, -\delta] ]) + (\sup_{x \in (-\delta, \delta)} f(x))(\text{sgn}[(-\delta, \delta)]) + (\sup_{x \in [\delta, 1]} f(x))(\text{sgn}[ [\delta, 1] ]) = 0 + (\sup_{x \in (-\delta, \delta)} f(x))2 + 0 = 2(\sup_{x \in (-\delta, \delta)} f(x)) \),根据(1),可得\( |(\sup_{x \in (-\delta, \delta)} f(x)) - f(0)| < \dfrac{\epsilon}{2} \),进而可得\( |2(\sup_{x \in (-\delta, \delta)} f(x)) - 2f(0)| < \epsilon \),即\( |p.c. \int_{[-1, 1]} g_{\epsilon} \mathop{} d \text{sgn} - 2f(0)| < \epsilon \),于是可得\( \overline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} \leq p.c. \int_{[-1, 1]} g_{\epsilon} \mathop{} d \text{sgn} \leq 2f(0) + \epsilon \) (2) 。同理可得\( \exists \)次要于\( f \)的分段常数函数\( h_{\epsilon}: [-1, 1] \to \mathbf{R} \) 满足\( |p.c. \int_{[-1, 1]} h_{\epsilon} \mathop{} d \text{sgn} - 2f(0)| < \epsilon \),于是可得\( \underline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} \geq p.c. \int_{[-1, 1]} h_{\epsilon} \mathop{} d \text{sgn} \geq 2f(0) - \epsilon \) (3) 。由(2)、(3)可得\( 2f(0) - \epsilon \leq \underline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} \leq \overline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} \leq 2f(0) + \epsilon \),进而可得\( |\underline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} - 2(f0)| < \epsilon \) 以及\( |\overline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} - 2(f0)| < \epsilon \),又\( \epsilon \)可以任意小,于是可得\( \underline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} = \overline{\int}_{[-1, 1]} f \mathop{} d \text{sgn} = 2f(0) \)。

综上,有\( f \)黎曼-斯蒂尔杰斯可积且\( \int_{[-1, 1]} f \mathop{} d \text{sgn} = 2f(0) \)。

证毕。

章节11.9

练习11.9.1

题目:

Let \( f: [0, 1] \to \mathbf{R} \) be the function in Exercise 9.8.5. Show that for every rational number \( q \in \mathbf{Q} \cap (0, 1) \), the function \( F: [0, 1] \to \mathbf{R} \) defined by the formula \( F(x) := \int_0^x f(y) \mathop{} dy \) is not differentiable at \( q \).

证明:

\( \forall q \in \mathbf{Q} \cap (0, 1) \),根据可微的定义,\( F \)在\( q \)可微当且仅当\( \lim_{x \to q; x \in [0, 1] \setminus \{ q \}} \dfrac{F(x) - F(q)}{x - q} \)存在,这需要其左右极限相等,即 \( \lim_{x \to q; x \in [0, q)} \dfrac{F(x) - F(q)}{x - q} = \lim_{x \to q; x \in (q, 1]} \dfrac{F(x) - F(q)}{x - q} \)。

针对左极限:由\( f \)单调递增以及定理11.4.1的5和6,我们可得 \( \lim_{x \to q; x \in [0, q)} \dfrac{F(x) - F(q)}{x - q} = \lim_{x \to q; x \in [0, q)} \dfrac{-(\int_x^q f(y) \mathop{} dy)}{x - q} = \lim_{x \to q; x \in [0, q)} \dfrac{\int_x^q f(y) \mathop{} dy}{q - x} \leq \lim_{x \to q; x \in [0, q)} \dfrac{\int_x^q f(q) \mathop{} dy}{q - x} = \dfrac{f(q)(q - x)}{q - x} = f(q) \)。

接下来我们考虑右极限:根据练习9.8.5第二部分的证明,可得\( \exists n \in \mathbf{N}, q = q(n) \) 且\( \forall x > q, f(x) \geq f(q) + 2^{-n} \),于是再由定理11.4.1的5和6,我们可得 \( \lim_{x \to q; x \in (q, 1]} \dfrac{F(x) - F(q)}{x - q} = \lim_{x \to q; x \in (q, 1]} \dfrac{\int_q^x f(y) \mathop{} dy}{x - q} \geq \lim_{x \to q; x \in (q, 1]} \dfrac{\int_q^x (f(q) + 2^{-n}) \mathop{} dy}{x - q} = \dfrac{(f(q) + 2^{-n})(x - q)}{x - q} = f(q) + 2^{-n} \)。

综上,可得\( |\lim_{x \to q; x \in [0, q)} \dfrac{F(x) - F(q)}{x - q} - \lim_{x \to q; x \in (q, 1]} \dfrac{F(x) - F(q)}{x - q}| \geq 2^{-n} \),于是有\( \lim_{x \to q; x \in [0, q)} \dfrac{F(x) - F(q)}{x - q} \neq \lim_{x \to q; x \in (q, 1]} \dfrac{F(x) - F(q)}{x - q} \),这意味着\( \lim_{x \to q; x \in [0, 1] \setminus \{ q \}} \dfrac{F(x) - F(q)}{x - q} \)不存在,即\( F \)在\( q \)不可微。

证毕。

练习11.9.2

题目:

Prove Lemma 11.9.5. (Hint: apply the mean-value theorem, Corollary 10.2.9 (or Proposition 10.3.3), to the function \( F - G \). One can also prove this lemma using the second Fundamental theorem of calculus (how?), but one has to be careful since we do not assume \( f \) to be Riemann integrable.)

Lemma 11.9.5的内容:

Let \( I \) be a bounded interval, and let \( f: I \to \mathbf{R} \) be a function. Let \( F: I \to \mathbf{R} \) and \( G: I \to \mathbf{R} \) be two antiderivatives of \( f \). Then there exists a real number \( C \) such that \( F(x) = G(x) + C \) for all \( x \in I \).

证明:

如果\( I = \empty \),则任取\( C \in \mathbf{R} \)均成立。

如果\( \exists a \in \mathbf{R}, I = \{ a \} \),则令\( C = F(a) - G(a) \),此时有\( \forall x \in I, x = a \),于是有\( F(x) = G(x) + (F(a) - G(a)) = G(x) + C \)。

如果\( I \)不为空集,也不为单例集,则\( \exists a < b \in \mathbf{R} \),使得\( I \)为\( [a, b], (a, b], (a, b), [a, b) \)形式的其中一种,这几种形式下均满足\( \forall x \in I, x \)为\( I \)的极限点,根据反导数的定义,这意味着\( \forall x \in I, F'(x) = f(x), G'(x) = f(x) \),除此之外,由\( F, G \)在\( I \)上可微,可得\( F, G \)在\( I \)上连续。

接下来我们分两个版本来证明。

使用推论10.2.9的版本:

任取\( x_0 \in I \),则\( \forall x < x_0 \in I \),我们有\( [x, x_0] \subseteq I \)且\( F, G \)在\( [x, x_0] \)上可微,于是有\( F - G \)在\( [x, x_0] \)上可微,根据推论10.2.9,可得\( \exists c \in [x, x_0], (F - G)'( c) = \dfrac{(F - G)(x_0) - (F - G)(x)}{x_0 - x} = F'( c) - G'( c) = f( c) - f( c) = 0 \),于是可得\( (F - G)(x_0) - (F - G)(x) = 0(x_0 - x) = 0 \),进而可得\( F(x) - G(x) = (F - G)(x) = (F - G)(x_0) \),令\( C = (F - G)(x_0) \),则有\( F(x) = G(x) + C \),除此之外,针对\( x_0 \),我们也有\( F(x_0) = G(x_0) + (F(x_0) - G(x_0)) = G(x_0) + C \),综上,可得\( \forall x \leq x_0 \in I \),\( F(x) = G(x) + C \),同理易证\( \forall x \geq x_0 \in I, F(x) = G(x) + C \),合起来,有\( \forall x \in I, F(x) = G(x) + C \)。

使用微积分第二基本定理的版本:

任取\( x_0 \in I \),则\( \forall x < x_0 \in I \),我们有\( [x, x_0] \subseteq I \)且\( F, G \)在\( [x, x_0] \)上可微,于是有\( F - G \)在\( [x, x_0] \)上可微。 \( \forall x \in I \),我们有\( (F - G)'(x) = F'(x) - G'(x) = f(x) - f(x) = 0 \),可得\( F - G \)是\( I \)上常数函数\( 0 \)的反导数,又\( I \)上的常数函数\( 0 \)黎曼可积,于是\( \forall x < x_0 \in I \),根据微积分第二基本定理,可得\( \int_{[x, x_0]} 0 = 0 = (F - G)(x_0) - (F - G)(x) \),进而可得\( F(x) - G(x) = (F - G)(x) = (F - G)(x_0) \),令\( C = (F - G)(x_0) \),则有\( F(x) = G(x) + C \),除此之外,针对\( x_0 \),我们也有\( F(x_0) = G(x_0) + (F(x_0) - G(x_0)) = G(x_0) + C \),综上,可得\( \forall x \leq x_0 \in I \),\( F(x) = G(x) + C \),同理易证\( \forall x \geq x_0 \in I, F(x) = G(x) + C \),合起来,有\( \forall x \in I, F(x) = G(x) + C \)。

证毕。

练习11.9.3

题目:

Let \( a < b \) be real numbers, and let \( f: [a, b] \to \mathbf{R} \) be a monotone increasing function. Let \( F: [a, b] \to \mathbf{R} \) be the function \( F(x) := \int_{[a, x]} f \). Let \( x_0 \) be an element of (a, b). Show that \( F \) is differentiable at \( x_0 \) if and only if \( f \) is continuous at \( x_0 \). (Hint: one direction is taken care of by one of the fundamental theorems of calculus. For the other, consider left and right limits of \( f \) and argue by contradiction.)

注:

参考了First Fundamental theorem of Calculus Corollary中Matthew Buck回答以及该回答下的讨论。

证明:

根据定理11.6.1,可得\( f \)黎曼可积。

必要性:

如果\( F \)在\( x_0 \)可微,此时假设\( f \)在\( x_0 \)不连续:

我们先证明\( f \)在\( x_0 \)点的左右极限存在,先考虑左极限,\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in ([a, b] \cap (-\infty, x_0) = [a, x_0)) \) 且\( \lim_{n \to \infty} a_n = x_0 \),我们打算证明\( \lim_{n \to \infty} f(a_n) \) 存在且\( = \sup_{x \in [a, x_0)} f(x) \),然后用定理9.3.9:由\( f \)单调递增,可得序列\( (f(a_n))_{n = 0}^{\infty} \)有上界\( f(x_0) \),又单调递增有上界的数列收敛于它的上确界,可得\( (f(a_n))_{n = 0}^{\infty} \)收敛于\( \sup(f(a_n))_{n = 0}^{\infty} \),我们假设\( \sup(f(a_n))_{n = 0}^{\infty} \neq \sup_{x \in [a, x_0)} f(x) \),这意味着\( \sup(f(a_n))_{n = 0}^{\infty} < \sup_{x \in [a, x_0)} f(x) \) (因为后者考虑的元素范围更广),于是\( \exists x_1 \in [a, x_0), f(x_1) > \sup(f(a_n))_{n = 0}^{\infty} \),而由\( f \)单调递增,这意味着\( \forall n \in \mathbf{N}, a_n < x_1 \),易证\( \lim_{n \to \infty} a_n \neq x_0 \),矛盾,于是假设不成立,有\( \sup(f(a_n))_{n = 0}^{\infty} = \sup_{x \in [a, x_0)} f(x) \)。综上,有\( \forall \)序列\( (a_n)_{n = 0}^{\infty} \)满足 \( \forall n \in \mathbf{N}, a_n \in ([a, b] \cap (-\infty, x_0) = [a, x_0)) \) 且\( \lim_{n \to \infty} a_n = x_0 \),我们均有 \( \lim_{n \to \infty} f(a_n) = \sup(f(a_n))_{n = 0}^{\infty} = \sup_{x \in [a, x_0)} f(x) \),根据定理9.3.9,可得 \( \lim_{x \to x_0^-; x \in [a, b]} f \)存在,同理易证\( \lim_{x \to x_0^+; x \in [a, b]} f \)存在。

令\( L^- := \lim_{x \to x_0^-; x \in [a, b]} f, L^+ := \lim_{x \to x_0^+; x \in [a, b]} f \),由\( f \)在\( x_0 \)不连续,可得\( L^- \neq f(x_0), L^+ \neq f(x_0) \),进而由\( f \)单调递增,可得\( L^- < f(x_0) < L^+ \)。令\( \epsilon_0 := \min(\dfrac{f(x_0) - L^-}{2}, \dfrac{L^+ - f(x_0)}{2}) \),则有\( L^- < f(x_0) - \epsilon_0 < f(x_0) < f(x_0) + \epsilon_0 < L^+ \)。由\( f \)单调递增,可得\( \forall x < x_0, f(x) \leq L^- < f(x_0) - \epsilon_0 \) 以及\( \forall x > x_0, f(x) \geq L^+ > f(x_0) + \epsilon_0 \),于是可得\( \forall x < x_0, F(x_0) - F(x) = \int_{[x, x_0]} f \leq \int_{[x, x_0]} (f(x_0) - \epsilon_0) = (f(x_0) - \epsilon_0)(x_0 - x) \) 以及\( \forall x > x_0, F(x) - F(x_0) = \int_{[x_0, x]} f \geq \int_{[x_0, x]} (f(x_0) + \epsilon_0) = (f(x_0) + \epsilon_0)(x - x_0) \),进而可得\( \lim_{x \to x_0^-; x \in [a, b]} \dfrac{F(x) - F(x_0)}{x - x_0} = \lim_{x \to x_0^-; x \in [a, b]} \dfrac{F(x_0) - F(x)}{x_0 - x} \leq \dfrac{(f(x_0) - \epsilon_0)(x_0 - x)}{x_0 - x} = f(x_0) - \epsilon_0 \) 以及\( \lim_{x \to x_0^+; x \in [a, b]} \dfrac{F(x) - F(x_0)}{x - x_0} \geq \dfrac{(f(x_0) + \epsilon_0)(x_0 - x)}{x_0 - x} = f(x_0) + \epsilon_0 \),这意味着\( \lim_{x \to x_0^+; x \in [a, b]} \dfrac{F(x) - F(x_0)}{x - x_0} \neq \lim_{x \to x_0^-; x \in [a, b]} \dfrac{F(x) - F(x_0)}{x - x_0} \),但这和\( F \)在\( x_0 \)可微矛盾,因此一开始的假设不成立,有\( f \)在\( x_0 \)连续。

充分性:

如果\( f \)在\( x_0 \)连续,此时根据微积分第一基本定理,可得\( F \)在\( x_0 \)可微。

证毕。

章节11.10

练习11.10.1

题目:

Prove Proposition 11.10.1. (Hint: first use Corollary 11.5.2 and Theorem 11.4.5 to show that \( F G' \) and \( F' G \) are Riemann integrable. Then use the product rule (Theorem 10.1.13(4)).)

Proposition 11.10.1的内容:

(Integration by parts formula). Let \( I = [a, b] \), and let \( F: [a, b] \to \mathbf{R} \) and \( G: [a, b] \to \mathbf{R} \) be differentiable functions on \( [a, b] \) such that \( F' \) and \( G' \) are Riemann integrable on \( I \). Then we have \( \int_{[a, b]} F G' = F(b)G(b) - F(a)G(a) - \int_{[a, b]} F' G \).

证明:

由定理10.1.10以及\( F, G \)在\( I \)上可微,可得\( F, G \)在\( I \)上连续,再根据推论11.5.2,可得\( F, G \)在\( I \)上黎曼可积,再根据定理11.4.5以及\( F', G' \) 在\( I \)上黎曼可积,可得\( F G', F' G \)在\( I \)上黎曼可积。根据微积分第二基本定理,可得 \( \forall x \in I, \int_{[a, b]} (FG)'(x) = (FG)(b) - (FG)(a) = F(b)G(b) - F(a)G(a) \) (1) (注:\( I \)为闭区间,集合内所有点均为极限点),根据定理10.1.13的4,可得\( \forall x \in I, (FG)'(x) = F'(x)G(x) + F(x)G'(x) \) (2) ,根据定理11.4.1的1以及(2),可得 \( \int_{[a, b]} (FG)' = \int_{[a, b]} (F' G + F G') = (\int_{[a, b]} F' G) + (\int_{[a, b]} F G') \),再结合(1),可得\( (\int_{[a, b]} F' G) + (\int_{[a, b]} F G') = F(b)G(b) - F(a)G(a) \),于是可得 \( \int_{[a, b]} F G' = F(b)G(b) - F(a)G(a) - \int_{[a, b]} F' G \)。

证毕。

练习11.10.2

题目:

Fill in the gaps marked (why?) in the proof of Lemma 11.10.5.

Lemma 11.10.5的内容:

(Change of variables formula I). Let \( [a, b] \) be a closed interval, and let \( \phi : [a, b] \to [\phi(a), \phi(b)] \) be a continuous monotone increasing function. Let \( f : [\phi(a), \phi(b)] \to \mathbf{R} \) be a piecewise constant function on \( [\phi(a), \phi(b)] \). Then \( f \circ \phi : [a, b] \to \mathbf{R} \) is also piecewise constant on \( [a, b] \), and \( \int_{[a, b]} f \circ \phi \mathop{} d \phi = \int_{[\phi(a), \phi(b)]} f \).

证明中引入的一些变量:

Let \( \mathbf{P} \) be a partition of \( [\phi(a), \phi(b)] \) such that \( f \) is piecewise constant with respect to \( \mathbf{P} \).

For each \( J \in \mathbf{P} \), let \( c_J \) be the constant value of \( f \) on \( J \).

第1个marked (why?)的内容:

For each interval \( J \in \mathbf{P} \), let \( \phi^{-1}(J) \) be the set \( \phi^{-1}(J) := \{ x \in [a, b] : \phi(x) \in J \} \). Then \( \phi^{-1}(J) \) is connected (why?).

证明:

由\( J \)是有界区间和引理11.1.4,可得\( J \)是连通的。

\( \forall x < y \in \phi^{-1}(J) \),我们需要证明\( [x, y] \subseteq \phi^{-1}(J) \),由\( x, y \in \phi^{-1}(J) \),可得\( x, y \in [a, b], \phi(x), \phi(y) \in J \),因为\( \phi \)单调递增,因此\( \phi(x) \leq \phi(y) \),针对\( \forall c \in [x, y] \),因为\( [x, y] \subseteq [a, b] \),因此\( c \in [a, b] \),接下来分类讨论:

  1. 如果\( \phi(x) = \phi(y) \),则由\( \phi \)单调递增,可得\( \phi(x) = \phi( c) = \phi(y) \in J \),又\( c \in [a, b] \),因此\( c \in \phi^{-1}(J) \)。
  2. 如果\( \phi(x) < \phi(y) \),则由\( \phi \)单调递增,可得\( \phi( c) \in [\phi(x), \phi(y)] \),由\( J \)是连通的,可得\( [\phi(x), \phi(y)] \subseteq J \),进而\( \phi( c) \in J \),又\( c \in [a, b] \),因此\( c \in \phi^{-1}(J) \)。

综上,\( \forall c \in [x, y] \),有\( c \in \phi^{-1}(J) \),即\( [x, y] \subseteq \phi^{-1}(J) \),因此\( \phi^{-1}(J) \)是连通的。

证毕。

第2个marked (why?)的内容:

Furthermore, \( c_J \) is the constant value of \( f \circ \phi \) on \( \phi^{-1}(J) \) (why?).

证明:

\( \forall x \in \phi^{-1}(J) \),有\( x \in [a, b], \phi(x) \in J \),特别的,由\( \phi(x) \in J \),可得\( f(\phi(x)) = c_J \),于是可得\( (f \circ \phi)(x) = f(\phi(x)) = c_J \)。

综上,\( c_J \)是\( f \circ \phi \)在\( \phi^{-1}(J) \)上的常数值。

证毕。

第3个marked (why?)的内容:

Define \( \mathbf{Q} := \{ \phi^{-1}(J) : J \in \mathbf{P} \} \) (ignoring the fact that \( \mathbf{Q} \) has been used to represent the rational numbers), then \( \mathbf{Q} \) partitions \( [a, b] \) (why?).

证明:

首先证明\( \bigcup_{K \in \mathbf{Q}} K = [a, b] \):\( \forall x \in \bigcup_{K \in \mathbf{Q}} K, \exists J \in \mathbf{P}, x \in \phi^{-1}(J) \),由\( \phi^{-1}(J) \)的定义,可得\( x \in [a, b] \),综上,有\( \bigcup_{K \in \mathbf{Q}} K \subseteq [a, b] \) (1) 。 \( \forall x \in [a, b], \exists J \in \mathbf{P}, \phi(x) \in J \),因此\( x \in \phi^{-1}(J) \),令\( K_J := \phi^{-1}(J) \),有\( K_J \subseteq \bigcup_{K \in \mathbf{Q}} K \),进而有\( x \in \bigcup_{K \in \mathbf{Q}} K \),综上,有\( [a, b] \subseteq \bigcup_{K \in \mathbf{Q}} K \),结合(1),可得\( \bigcup_{K \in \mathbf{Q}} K = [a, b] \)。

最后证明\( \forall C, D \in \mathbf{Q}, C \cap D = \empty \):假设\( \exists C, D \in \mathbf{Q}, C \cap D \neq \empty \),于是\( \exists x \in C \cap D \),由\( C, D \in \mathbf{Q} \),可得\( \exists J_C, J_D \in \mathbf{P}, C = \phi^{-1}(J_C), D = \phi^{-1}(J_D) \),于是有\( x \in \phi^{-1}(J_C) \cap \phi^{-1}(J_D) \),进而有\( \phi(x) \in J_C \cap J_D \),这和\( \mathbf{P} \)是\( [\phi(a), \phi(b)] \)的分区矛盾,因此假设不成立,有\( \forall C, D \in \mathbf{Q}, C \cap D = \empty \)。

综上,可得\( \mathbf{Q} \)是\( [a, b] \)的分区。

证毕。

第4个marked (why?)的内容:

And \( f \circ \phi \) is piecewise constant with respect to \( \mathbf{Q} \) (why?).

证明:

\( \forall K \in \mathbf{Q}, \exists J \in \mathbf{P}, K = \phi^{-1}(J) \),根据第2个marked (why?)的内容,可得\( c_J \)是\( f \circ \phi \)在\( K \)上的常数值,特别的,这意味着\( f \circ \phi \)在\( K \)上为常数函数。加上\( \mathbf{Q} \)是\( [a, b] \)的分区,可得\( f \circ \phi \)针对\( \mathbf{Q} \)为分段常数函数。

证毕。

第5个marked (why?)的内容:

But \( \phi[\phi^{-1}(J)] = |J| \) (why?).

证明:

由\( \phi^{-1}(J) \)是有界区间,可得\( \exists c < d \in \mathbf{R} \),使得\( \phi^{-1}(J) \)为\( [c, d], (c, d], (c, d), [c, d) \)形式的其中一种,又\( \phi \)为连续函数,因此\( \phi[\phi^{-1}(J)] = \phi(d) - \phi( c) \),而\( \phi \)单调递增,因此\( \phi( c) \)为\( J \)的左端点, \( \phi(d) \)为\( J \)的右端点,于是有\( |J| = \phi(d) - \phi( c) \),综上,有\( \phi[\phi^{-1}(J)] = |J| \)。

证毕。

练习11.10.3

题目:

Let \( a < b \) be real numbers, and let \( f : [a, b] \to \mathbf{R} \) be a Riemann integrable function. Let \( g : [-b, -a] \to \mathbf{R} \) be defined by \( g(x) := f(-x) \). Show that \( g \) is also Riemann integrable, and \( \int_{[-b, -a]} g = \int_{[a,b]} f \).

证明:

\( \forall \epsilon > 0 \):

因为\( (\int_{[a, b]} f) + \epsilon > \int_{[a, b]} f \),因此\( \exists [a, b] \)上主要于\( f \)的分段常数函数\( f_u \)使得 \( \int_{[a, b]} f_u < (\int_{[a, b]} f) + \epsilon \),同理可得 \( \exists [a, b] \)上次要于\( f \)的分段常数函数\( f_l \)使得 \( (\int_{[a, b]} f) - \epsilon < \int_{[a, b]} f_l \),又\( \int_{[a, b]} f_l \leq \int_{[a, b]} f \leq \int_{[a, b]} f_u \),因此可得\( (\int_{[a, b]} f) - \epsilon < \int_{[a, b]} f_l \leq \int_{[a, b]} f_u < (\int_{[a, b]} f) + \epsilon \) (1) 。由\( f_u, f_l \)为\( [a, b] \)上的分段常数函数,可得 \( \exists [a, b] \)的分区\( \mathbf{P_u}, \mathbf{P_l} \)使得 \( f_u \)针对\( \mathbf{P_u} \)为分段常数函数、\( f_l \)针对 \( \mathbf{P_l} \)为分段常数函数,令\( \mathbf{P} := \mathbf{P_u} \# \mathbf{P_l} \),根据引理11.1.18和引理11.2.7,可得\( f_u, f_l \)均针对\( \mathbf{P} \)为分段常数函数。

定义函数\( g_u : [-b, -a] \to \mathbf{R}, g_l : [-b, -a] \to \mathbf{R}, \forall x \in [-b, -a] \),令\( g_u := f_u(-x), g_l := f_l(-x) \),易得\( g_u \)为\( [-b, -a] \)上主要于\( g \)的分段常数函数、 \( g_l \)为\( [-b, -a] \)上次要于\( g \)的分段常数函数 (2) ,令\( \forall J \in \mathbf{P} \),令\( K_J := -J \),令\( \mathbf{P_g} := \bigcup_{J \in \mathbf{P}} \{ K_J \} \),易得\( \mathbf{P_g} \)为\( [-b, -a] \)的分区且 \( \forall J \in \mathbf{P} \),\( f_u \)在\( J \)上的常数值\( c_{uJ} = \) \( g_u \)在\( K_J \)上的常量值、\( f_l \)在\( J \)上的常数值\( c_{lJ} = \) \( g_l \)在\( K_J \)上的常量值,除此之外,还有\( |K_J| = |J| \),于是可得\( \int_{[-b, -a]} g_u = p.c. \int_{[\mathbf{P_g}]} g_u = \sum_{J \in \mathbf{P}} c_{uJ}|J| = p.c. \int_{[\mathbf{P}]} f_u = \int_{[a, b]} f_u, \int_{[-b, -a]} g_l = p.c. \int_{[\mathbf{P_g}]} g_l = \sum_{J \in \mathbf{P}} c_{lJ}|J| = p.c. \int_{[\mathbf{P}]} f_l = \int_{[a, b]} f_l \),再结合(1)和(2),可得\( (\int_{[a, b]} f) - \epsilon < \int_{[-b, -a]} g_l \leq \underline{\int}_{[-b, -a]} g \leq \overline{\int}_{[-b, -a]} g \leq \int_{[-b, -a]} g_u < (\int_{[a, b]} f) + \epsilon \),又该不等式对任意\( \epsilon \)都成立,因此可得 \( \underline{\int}_{[-b, -a]} g = \overline{\int}_{[-b, -a]} g = \int_{[a, b]} f \),即\( g \)黎曼可积且\( \int_{[-b, -a]} g = \int_{[a, b]} f \)。

证毕。

练习11.10.4

题目:

What is the analogue of Proposition 11.10.7 when \( \phi \) is monotone decreasing instead of monotone increasing? (When \( \phi \) is neither monotone increasing or monotone decreasing, the situation becomes significantly more complicated.)

Proposition 11.10.7的内容:

(Change of variables formula III). Let \( [a, b] \) be a closed interval, and let \( \phi : [a, b] \to [\phi(a), \phi(b)] \) be a differentiable monotone increasing function such that \( \phi' \) is Riemann integrable. Let \( f : [\phi(a), \phi(b)] \to \mathbf{R} \) be a Riemann integrable function on \( [\phi(a), \phi(b)] \). Then \( (f \circ \phi)\phi' : [a, b] \to \mathbf{R} \) is Riemann integrable on \( [a, b] \), and \( \int_{[a , b]} (f \circ \phi)\phi' = \int_{[\phi(a), \phi(b)]} f \).

Proposition 11.10.7对应的\( \phi \)改成单调递减函数的版本:

Let \( [a, b] \) be a closed interval, and let \( \phi : [a, b] \to [\phi(b), \phi(a)] \) be a differentiable monotone decreasing function such that \( \phi' \) is Riemann integrable. Let \( f : [\phi(b), \phi(a)] \to \mathbf{R} \) be a Riemann integrable function on \( [\phi(b), \phi(a)] \). Then \( (f \circ \phi)\phi' : [a, b] \to \mathbf{R} \) is Riemann integrable on \( [a, b] \), and \( \int_{[a , b]} (f \circ \phi)\phi' = -(\int_{[\phi(b), \phi(a)]} f) \).

理由:

定理11.10.7是通过定理11.10.6以及推论11.10.3证明的,故我们考虑下这两个定理在\( \phi \)改成单调递减函数会怎么样?

先看推论11.10.3,它的证明依赖于定理11.10.2,观察定理11.10.2的证明,可以发现在\( \alpha \)改成单调递减函数后,定理11.10.2的结论以及证明均不变。回来继续考虑推论11.10.3,在\( \alpha \)改成单调递减函数后(加上\( \alpha \)可微), \( \alpha' \)会变成负数,于是从“Since \( \alpha' \) is non-negative …”开始要改成“Since \( \alpha' \) is negative and \( \underline{f} \) minorizes \( f \), then \( \underline{f} \alpha' \) majorizes \( f \alpha' \). Thus \( \underline{\int}_{[a, b]} \underline{f} \alpha' = \overline{\int}_{[a, b]} \underline{f} \alpha' \geq \overline{\int}_{[a, b]} f \alpha' \). Thus \( \overline{\int}_{[a, b]} f \alpha' \leq (\int_{[a, b]} f \mathop{} d \alpha) + \epsilon \). Similarly we have \( (\int_{[a, b]} f \mathop{} d \alpha) - \epsilon \leq \underline{\int}_{[a, b]} f \alpha' \). Since these statements are true for any \( \epsilon > 0 \), we must have \( \int_{[a, b]} f \mathop{} d \alpha \leq \underline{\int}_{[a, b]} f \alpha' \leq \overline{\int}_{[a, b]} f \alpha' \leq \int_{[a, b]}) f \mathop{} d \alpha \) and the claim follows.”会发现结论还是一样的。

接着看定理11.10.6,它的证明依赖于引理11.10.5,我们先看引理11.10.5,在\( \phi \)改成连续单调递减函数后,首先所有\( [\phi(a), \phi(b)] \) 需要改成\( [\phi(b), \phi(a)] \),针对第1个marked (why?)的内容,虽然证明方法不一样了,但是结论仍然相同,针对第2、3、4个marked (why?)的内容,它们的证明不变,针对第5个marked (why?)的内容,由于\( \phi \)为单调递减函数,因此结论会变成\( \phi[\phi^{-1}(J)] = -|J| \),于是引理11.10.5的最终结论会变成 \( \int_{[a, b]} f \circ \phi \mathop{} d \phi = -(\int_{[\phi(b), \phi(a)]} f) \)。回来继续考虑定理11.10.6,在\( \phi \)改成连续单调递减函数后,一样的,所有\( [\phi(a), \phi(b)] \)需要改成\( [\phi(b), \phi(a)] \),从“Applying Lemma 11.10.5, we obtain …”开始要改成“\( (\int_{[\phi(b), \phi(a)]} f) - \epsilon \leq -(\int_{[a, b]} \underline{f} \circ \phi \mathop{} d \phi) \leq -(\int_{[a, b]} \overline{f} \circ \phi \mathop{} d \phi) \leq (\int_{[\phi(b), \phi(a)]} f) + \epsilon \). Since \( \underline{f} \circ \phi \) piecewise constant and minorizes \( f \circ \phi \), we have \( -(\int_{[a, b]} \underline{f} \circ \phi \mathop{} d \phi) \geq -(\underline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \) while similarly we have \( -(\int_{[a, b]} \overline{f} \circ \phi \mathop{} d \phi) \leq -(\overline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \). Thus \( (\int_{[\phi(b), \phi(a)]} f) - \epsilon \leq -(\overline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \leq -(\underline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \leq (\int_{[\phi(b), \phi(a)]} f) + \epsilon \). Since \( \epsilon > 0 \) was arbitrary, this implies that \( \int_{[\phi(b), \phi(a)]} f \leq -(\overline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \leq -(\underline{\int}_{[a, b]} f \circ \phi \mathop{} d \phi) \leq \int_{[\phi(b), \phi(a)]} f \)”,于是最终结论变成\( \int_{[\phi(b), \phi(a)]} f = -(\int_{[a, b]} f \circ \phi \mathop{} d \phi) \),或者说是\( \int_{[a, b]} f \circ \phi \mathop{} d \phi = -(\int_{[\phi(b), \phi(a)]} f) \)。

最后回来考虑定理11.10.7,在\( \phi \)改成连续单调递减函数后,一样的,首先所有\( [\phi(a), \phi(b)] \)需要改成\( [\phi(b), \phi(a)] \),最后的结论变成\( \int_{[a , b]} (f \circ \phi)\phi' = -(\int_{[\phi(b), \phi(a)]} f) \)。

参考文章

  1. First Fundamental theorem of Calculus Corollary中Matthew Buck回答以及该回答下的讨论。