目录

陶哲轩Analysis I习题的参考解答及思考(附录B)

附录B

版本

Analysis I(第3版)。

章节B.1

练习B.1.1

题目:

The purpose of this exercise is to demonstrate that the procedure of long addition taught to you in elementary school is actually valid. Let \( A = a_n \dots a_0 \) and \( B = b_m \dots b_0 \) be positive integer decimals. Let us adopt the convention that \( a_i = 0 \) when \( i > n \), and \( b_i = 0 \) when \( i > m \); for instance, if \( A = 372 \), then \( a_0 = 2, a_1 = 7, a_2 = 3, a_3 = 0, a_4 = 0 \), and so forth. Define the numbers \( c_0, c_1, \dots \) and \( \epsilon_0, \epsilon_1 \dots \) recursively by the following long addition algorithm.

  1. We set \( \epsilon_0 := 0 \).
  2. Now suppose that \( \epsilon_i \) has already been defined fo some \( i \geq 0 \). If \( a_i + b_i + \epsilon_i < 10 \), we set \( c_i := a_i + b_i + \epsilon_i \) and \( \epsilon_{i + 1} := 0 \); otherwise, if \( a_i + b_i + \epsilon_i \geq 10 \), we set \( c_i := a_i + b_i + \epsilon_i -10 \) and \( \epsilon_{i + 1} := 1 \). (The number \( \epsilon_{i + 1} \) is the “carry digit” from the \( i^{th} \) decimal place to the \( (i + 1)^{th} \) decimal place.)

Prove that the numbers \( c_0, c_1, \dots \) are all digits, and that there exists an \( l \) such that \( c_l \neq 0 \) and \( c_i = 0 \) for all \( i > l \). Then show that \( c_l c_{l - 1} \dots c_1 c_0 \) is the decimal representation of \( A + B \).

Note that one could in fact use this algorithm to define addition, but it would look extremely complicated, and to prove even such simple facts as \( (a + b) + c = a + (b + c) \) would be rather difficult. This is one of the reasons why we have avoided the decimal system in our construction of the natural numbers. The procedure for long multiplication (or long subtraction, or long division) is even worse to lay out rigorously; we will not do so here.

证明:

先证明\( c_0, c_1, \dots \)都是数字: \( \forall i \geq 0 \),如果\( a_i + b_i + \epsilon_i < 10 \),则\( 0 \leq c_i = a_i + b_i + \epsilon_i < 10 \),即\( c_i \)为数字,如果\( a_i + b_i + \epsilon_i \geq 10 \),则由\( a_i, b_i \)为数字以及\( \epsilon_i = 0 \)或\( \epsilon_i = 1 \),可得\( 10 \leq a_i + b_i + \epsilon_i \leq 9 + 9 + 1 = 19 \),于是\( 0 \leq c_i = a_i + b_i + \epsilon_i - 10 = a_i + b_i - 10 \leq 9 \),即\( c_i \)为数字。

接着证明\( \exists l \in \mathbf{N}, c_l \neq 0, \forall i > l, c_i = 0 \):令\( p := \max(n, m) \),如果\( a_p + b_p + \epsilon_p < 10 \),则\( c_p = a_p + b_p + \epsilon_p \neq 0 \),而\( \forall i > p, c_i = a_i + b_i + \epsilon_i = 0 + 0 + 0 = 0 \),如果\( a_p + b_p + \epsilon_p \geq 10 \),则\( \epsilon_{p + 1} = 1 \),进而\( a_{p + 1} + b_{p + 1} + \epsilon_{p + 1} = 0 + 0 + 1 = 1 \neq 0 \),但\( a_{p + 1} + b_{p + 1} + \epsilon_{p + 1} = 1 < 10 \),因此 \( \forall i > p + 1 \),有\( c_i = a_i + b_i + \epsilon_i = 0 + 0 + 0 = 0 \),综上,\( \exists l \in \mathbf{N}, c_l \neq 0, \forall i > l, c_i = 0 \)。根据上面的证明,可得\( l = p \)或\( l = p + 1 \),前者在\( a_p + b_p + \epsilon_p < 10 \)取到,此时\( c_l = a_p + b_p + \epsilon_p \),后者在\( a_p + b_p + \epsilon_p \geq 10 \)时取到,此时\( c_l = 1 \)。

最后,证明\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示:考虑\( n \geq m \)的情况,\( m > n \)的情况类似,这里不考虑,令\( p := \max(n, m) = n \),归纳假设\( \forall p' < p \),命题均成立,我们证明命题对\( p \)成立:

  1. 如果\( p = 0 \),则\( A = a_0, B = b_0 \):
    1. 如果\( l = p = 0 \),则\( a_0 + b_0 + \epsilon_0 = a_0 + b_0 < 10, c_l c_{l - 1} \dots c_1 c_0 = c_0 = a_0 + b_0 + \epsilon_0 \) 明显是\( A + B \)的十进制表示。
    2. 如果\( l = p + 1 = 1 \),则\( a_0 + b_0 + \epsilon_0 = a_0 + b_0 \geq 10, c_l c_{l - 1} \dots c_1 c_0 = c_1c_0 \),其中\( c_1 = 1, c_0 = a_0 + b_0 + \epsilon_0 - 10 \),于是可得\( c_1 c_0 = (1 \times 10) + (a_0 + b_0 + \epsilon_0 - 10) = a_0 + b_0 = A + B \),即\( c_l c_{l - 1} \dots c_1 c_0 = c_1 c_0 \)是\( A + B \)的十进制表示。
  2. 如果\( p > 0 \),则令\( C := a_{n - 1} \dots a_0 \):如果\( n = m \),则令\( D := b_{m - 1} \dots b_0 \),令\( p' := \max(n - 1, m - 1) = n - 1 \) 如果\( n > m \),则令\( D := b_m \dots b_0 = B \),令\( p' := \max(n - 1, m) = n - 1 \),不管什么情况,均有\( p' < p \),根据归纳假设\( c_{l'} c_{l' - 1} \dots c_1 c_0 \)为\( C + D \)的十进制表示( 下面的证明中,注意下 ,有时候\( D \)和\( B \)相差一个最高位,有时候\( D = B \),注意两种情况都验证下等式什么的是否成立):
    1. 如果\( l' = p' = n - 1 \),则\( c_{l'} = a_{p'} + b_{p'} + \epsilon_{p'} < 10 \),此时\( \epsilon_p = 0 \):
      1. 如果\( l = p = n \),则\( a_p + b_p + \epsilon_p < 10 \),此时\( c_l = a_p + b_p + \epsilon_p, c_l c_{l - 1} \dots c_1 c_0 = c_l c_{l'} \dots c_1 c_0 = (c_l \times (10)^l) + (c_{l'} c_{l' - 1} \dots c_1 c_0) = ((a_p + b_p + \epsilon_p) \times (10)^l) + (C + D) = ((a_p + b_p) \times (10)^l) + (C + D) \),而\( A + B = (a_p \times (10)^l) + (b_p \times (10)^l) + C + D = ((a_p + b_p) \times (10)^l) + (C + D) = c_l c_{l - 1} \dots c_1 c_0 \),即\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示。
      2. 如果\( l = p + 1 = n + 1 \),则\( a_p + b_p + \epsilon_p \geq 10 \),此时\( c_l = 1, c_{l - 1} = a_p + b_p + \epsilon_p - 10, c_l c_{l - 1} \dots c_1 c_0 = c_l c_{l - 1} c_{l'} \dots c_1 c_0 = (c_l \times (10)^l) + (c_{l - 1} \times (10)^{l - 1}) + (c_{l'} c_{l' - 1} \dots c_1 c_0) = 10^l + ((a_p + b_p + \epsilon_p - 10) \times (10)^{l - 1}) + (C + D) \),而\( A + B = (a_p \times (10)^{l - 1}) + (b_p \times (10)^{l - 1}) + C + D = ((a_p + b_p) \times (10)^{l - 1}) + (C + D) = ((a_p + b_p + \epsilon_p) \times (10)^{l - 1}) + (C + D) \text{(注:} \epsilon_p = 0 \text{)} = (10 \times (10)^{l - 1}) + ((a_p + b_p + \epsilon_p - 10) \times (10)^{l - 1}) + (C + D) = 10^l + ((a_p + b_p + \epsilon_p - 10) \times (10)^{l - 1}) + (C + D) = c_l c_{l - 1} \dots c_1 c_0 \),即\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示。
    2. 如果\( l' = p' + 1 = n \),则\( a_{p'} + b_{p'} + \epsilon_{p'} \geq 10 \),此时\( c_{l'} = 1, c_{l' - 1} = a_{p'} + b_{p'} + \epsilon_{p'} - 10, \epsilon_p = \epsilon_{l'} = c_{l'} = 1 \):
      1. 如果\( l = p = n \),则\( a_p + b_p + \epsilon_p < 10 \),此时\( c_l = a_p + b_p + \epsilon_p, c_l c_{l - 1} \dots c_1 c_0 = (c_l \times (10)^l) + (c_{l - 1} c_{l - 2} \dots c_1 c_0) = (c_l \times (10)^l) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = ((a_p + b_p + \epsilon_p) \times (10)^l) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = ((a_p + b_p) \times (10)^l) + (\epsilon_p \times 10^l) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = ((a_p + b_p) \times (10)^l) + (c_{l'} \times 10^{l'}) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = ((a_p + b_p) \times (10)^l) + (c_{l'} c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = ((a_p + b_p) \times (10)^l) + (C + D) \),而\( A + B = (a_p \times (10)^l) + (b_p \times (10)^l) + C + D = ((a_p + b_p) \times (10)^l) + (C + D) = c_l c_{l - 1} \dots c_1 c_0 \),即\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示。
      2. 如果\( l = p + 1 = n + 1 \),则\( a_p + b_p + \epsilon_p \geq 10 \),此时\( c_l = 1, c_{l - 1} = a_p + b_p + \epsilon_p - 10, c_l c_{l - 1} \dots c_1 c_0 = c_l c_{l - 1} \dots c_1 c_0 = (c_l \times (10)^l) + (c_{l - 1} \times (10)^{l - 1}) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = 10^l + ((a_p + b_p + \epsilon_p - 10) \times (10)^{l - 1}) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = 10^l + ((a_p + b_p - 10) \times (10)^{l - 1}) + (\epsilon_p \times (10)^{l - 1}) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = 10^l + ((a_p + b_p - 10) \times (10)^{l - 1}) + (c_{l'} \times (10)^{l'}) + (c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = 10^l + ((a_p + b_p - 10) \times (10)^{l - 1}) + (c_{l'} c_{l' - 1} c_{l' - 2} \dots c_1 c_0) = 10^l + ((a_p + b_p - 10) \times (10)^{l - 1}) + (C + D) \),而\( A + B = (a_p \times (10)^{l - 1}) + (b_p \times (10)^{l - 1}) + C + D = ((a_p + b_p) \times (10)^{l - 1}) + (C + D) = (10 \times (10)^{l - 1}) + ((a_p + b_p - 10) \times (10)^{l - 1}) + (C + D) = 10^l + ((a_p + b_p - 10) \times (10)^{l - 1}) + (C + D) = c_l c_{l - 1} \dots c_1 c_0 \),即\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示。

综上,命题对\( p \)成立,归纳完毕,命题成立,有\( c_l c_{l - 1} \dots c_1 c_0 \)是\( A + B \)的十进制表示。

证毕。

章节B.2

练习B.2.1

题目:

If \( a_n \dots a_0 . a_{-1} a_{-2} \dots \) is a real decimal, show that the series \( \sum_{i = -\infty}^{n} a_i \times 10^i \) is absolutely convergent.

证明:

由于\( \forall i \leq n, a_i \geq 0 \),因此证明 \( \sum_{i = -\infty}^{n} a_i \times 10^i \)绝对收敛,我们仅需要证明\( \sum_{i = -\infty}^{n} a_i \times 10^i \)收敛即可。

严格来讲,\( \sum_{i = -\infty}^{n} a_i \times 10^i \)是未定义的,第7章对级数的定义中,我们均要求下标下限\( m \)是整数,而不能是\( -\infty \),这里将\( \sum_{i = -\infty}^{n} a_i \times 10^i \)理解为 \( (\sum_{i = 0}^{n} a_i \times 10^i) + (\sum_{i = 1}^{+\infty} a_i \times 10^{-i}) \),即一个有限级数和一个无限级数的和,当然,这个无限级数要收敛,这个和才有意义。

我们先证明\( \sum_{i = 1}^{+\infty} a_i \times 10^{-i} \)收敛,用定理7.3.1来证明,我们打算证明\( \forall N \geq 1 \),均有\( \sum_{n = 1}^{N} a_i \times 10^{-i} \leq 1 - 10^{-N} \),使用数学归纳法证明:当\( N = 1 \)时,\( \sum_{n = 1}^{N} a_i \times 10^{-i} = a_1 \times 10^{-1} \),由于\( a_1 \)是数字,因此\( 0 \leq a_1 \leq 9 \),进而\( 0 \leq a_1 \times 10^{-1} \leq 9 \times 10^{-1} = (10 \times 10^{-1}) - (10^{-1}) = 1 - 10^{-1} \),因此\( \sum_{n = 1}^{N} a_i \times 10^{-i} \leq 1 - 10^{-N} \),\( N = 1 \)时成立。归纳假设当\( N = K \)时成立,当\( N = K + 1 \)时,根据归纳假设,\( \sum_{n = 1}^{K} a_i \times 10^{-i} \leq 1 - 10^{-K} \),进而\( \sum_{n = 1}^{K + 1} a_i \times 10^{-i} = (\sum_{n = 1}^{K} a_i \times 10^{-i}) + (a_{K + 1} \times 10^{-(K + 1)}) \leq (1 - 10^{-K}) + (a_{K + 1} \times 10^{-(K + 1)}) \),由于\( a_{K + 1} \)是数字,因此 \( 0 \leq a_{K + 1} \times 10^{-(K + 1)} \leq 9 \times 10^{-(K + 1)} \),于是有\( \sum_{n = 1}^{K + 1} a_i \times 10^{-i} \leq (1 - 10^{-K}) + (a_{K + 1} \times 10^{-(K + 1)}) \leq (1 - 10^{-K}) + (9 \times 10^{-(K + 1)}) = 1 - (10 \times 10^{-(K + 1)}) + (9 \times 10^{-(K + 1)}) = 1 - 10^{-(K + 1)} \),故\( N = K + 1 \)时也成立,归纳完毕,有\( \forall N \geq 1, \sum_{n = 1}^{N} a_i \times 10^{-i} \leq 1 - 10^{-N} \leq 1 \),根据定理7.3.1,可得\( \sum_{i = 1}^{+\infty} a_i \times 10^{-i} \)收敛,进而和\( (\sum_{i = 0}^{n} a_i \times 10^i) + (\sum_{i = 1}^{+\infty} a_i \times 10^{-i}) \) 是有意义的,或者说\( \sum_{i = -\infty}^{n} a_i \times 10^i \)绝对收敛。

证毕。

附注

由于十进制表示对应的级数是绝对收敛的,因此根据定理7.4.3,我们可以任意改变级数项相加的顺序,改变顺序后的级数仍然绝对收敛且级数的和相同,下面练习的证明中,我们可能会隐式用到该性质,届时不会明确说明用了定理7.4.3。

练习B.2.2

题目:

Show that the only decimal representations \( 1 = \pm a_n \dots a_0 . a_{-1} a_{-2} \dots \) of \( 1 \) are \( 1 = 1.000 \dots \) and \( 1 = 0.999 \dots \).

注:

练习B.2.3的证明不依赖于本练习,可以先证完练习B.2.3,然后再用练习B.2.3 来证明本练习,证明将非常简短,但是单纯靠练习B.2.3,我们只能知道\( 1 \) 有且仅有两个十进制表示,并不能知道这两个表示分别是\( 1.000 \dots, 0.999 \dots \),除非你引用练习B.2.3中两个十进制表示的构造过程(或者改变练习B.2.3的结论,直接连两个十进制表示也给出来)。

这里选择先证明本练习,证明本练习的成果,将为练习B.2.3的证明提供思路。

证明:

\( \forall 1 \)的十进制表示\( \pm a_n \dots a_0 . a_{-1} a_{-2} \dots \),由\( 1 > 0 \),可得十进制表示得取正号,不能取负号,即只能取\( a_n \dots a_0 . a_{-1} a_{-2} \dots = \sum_{i = -\infty}^{n} a_i \times 10^i \)。针对\( n \),如果\( n > 0 \),则\( \sum_{i = -\infty}^{n} a_i \times 10^i \geq \sum_{i = n}^{n} a_i \times 10^i = a_n \times 10^n > 1 \) (注:当\( n > 0 \)时,\( a_n \neq 0 \),只能\( > 0 \),这是十进制表示法的约定,只有\( n = 0 \)时,\( a_n \)才能取\( 0 \),否则整数就不会有唯一的十进制表示),这和\( a_n \dots a_0 . a_{-1} a_{-2} \dots \)是\( 1 \)的十进制表示矛盾,于是\( n \)必须\( = 0 \),即小数点左边必须有且仅有一位数字\( a_0 \)。我们考虑\( a_0 \)的值,如果\( a_0 > 1 \),则\( \sum_{i = -\infty}^{n} a_i \times 10^i \geq \sum_{i = n}^{n} a_i \times 10^i = a_n \times 10^n = a_0 \times 10^0 > 1 \times 10^0 = 1 \),这和\( a_n \dots a_0 . a_{-1} a_{-2} \dots \)是\( 1 \)的十进制表示矛盾,于是\( a_0 \leq 1 \),即\( a_0 = 1 \)或\( a_0 = 0 \):

  1. 如果\( a_0 = 1 \),则此时假设小数点右边存在不为\( 0 \)的数字,即\( \exists i_0 < 0, a_{i_0} > 0 \),则此时 \( \sum_{i = -\infty}^{n} a_i \times 10^i \geq \sum_{i = i_0}^{n} a_i \times 10^i \geq 1 + (a_{i_0} \times 10^{i_0}) > 1 \),这和\( a_n \dots a_0 . a_{-1} a_{-2} \dots \)是\( 1 \)的十进制表示矛盾,于是假设不成立,有\( \forall i < 0, a_i = 0 \),综上可得,\( a_0 = 1 \)的情况下, \( 1 \)的十进制表示只能是\( 1.000 \dots \)。
  2. 如果\( a_0 = 0 \),则此时假设小数点右边存在不为\( 9 \)的数字,即\( \exists i_0 < 0, 0 \leq a_{i_0} < 9 \),此时构造十进制表示 \( b_n \dots b_0 . b_{-1} b_{-2} \dots, \forall i \leq n \),如果\( i_0 < i \leq n \),则令\( b_i = a_i \),如果\( i = i_0 \),则令\( b_i = 9 \),如果\( i < i_0 \),则令\( b_i = 0 \),即拷贝\( i_0 \)左侧的所有数字,然后令\( i_0 \)位的数字为\( > a_{i_0} \)的数字\( 9 \),最后\( i_0 \)右侧的数字全部为\( 0 \) (具体的,构造的十进制表示形如\( 0.999900 \dots \)),可得\( \sum_{i = -\infty}^{n} a_i \times 10^i \leq \sum_{i = -\infty}^{n} b_i \times 10^i = \sum_{i = i_0}^{n} b_i \times 10^i < 1 \),这和\( a_n \dots a_0 . a_{-1} a_{-2} \dots \)是\( 1 \)的十进制表示矛盾,于是假设不成立,有\( \forall i < 0, a_i = 9 \),综上可得,\( a_0 = 0 \)的情况下, \( 1 \)的十进制表示只能是\( 0.999 \dots \)。

综上,\( 1 \)有且仅有两个十进制表示\( 1.000 \dots, 0.999 \dots \)。

证毕。

练习B.2.3

题目:

A real number \( x \) is said to be a terminating decimal if we have \( x = n / 10^{-m} \) for some integers \( n, m \). Show that if \( x \) is a non-zero terminating decimal, then \( x \) has exactly two decimal representations, while if \( x \) is not a non-zero terminating decimal, then \( x \) has exactly one decimal representation.

注:

严格来讲,下面的证明中,不同地方的\( \pm \)符号,除非明确讲明它们的正负号之间的关系,否则它们的正负号是独立的,不过通过上下文,读者很容易判断几个\( \pm \)符号的正负号之间的关系,于是下面证明中,为了简短点,一般不会明说对应的关系。

证明:

\( \forall x \in \mathbf{R} \):

如果\( x \)是非\( 0 \)有尽十进制:则\( \exists n, m \in \mathbf{Z}, x = n / 10^{-m} \),由\( x \neq 0 \),可得\( n \neq 0 \) (即类似\( 1.000 \dots \)这种可以通过有限个十进制数字表示的实数,比如,\( 1.0 \)、\( 1 \),明显这类实数可以通过一个整数(整数可以通过有限个十进制数字表示)整除\( 10^{-m} \)以右移来得到):我们先证明\( x \)至少有两个不同的十进制表示,根据定理B.1.4,我们知道\( n \)存在一个十进制表示,该十进制表示的小数点右侧数字全为\( 0 \),令\( \pm c_k \dots c_0 . c_{-1} c_{-2} \dots = \pm \sum_{i = -\infty}^{k} c_i \times 10^i \)为\( n \)的该十进制表示,可得\( k \geq 0, c_k > 0, \forall i < 0, c_i = 0 \),而\( x = n / 10^{-m} = (\pm \sum_{i = -\infty}^{k} c_i \times 10^i) / 10^{-m} = \pm \sum_{i = -\infty}^{k} c_i \times 10^{i + m} = \pm \sum_{i = -\infty}^{k + m} c_{i - m} \times 10^i \),构造新的十进制表示\( \pm a_{k + m} \dots a_0 . a_{-1} a_{-2} \dots, \forall i \leq k + m \),令\( a_i := c_{i - m} \),这里我们想说\( \pm a_{k + m} \dots a_0 . a_{-1} a_{-2} \dots = \pm \sum_{i = -\infty}^{k + m} a_i \times 10^i \)为\( x \)的一个十进制表示,但是有一个问题,\( k + m \)可能\( < 0 \),这是不允许的,小数点左边必须至少有一个数字,因此我们令\( k_2 := \max(k + m, 0), \forall k + m < i \leq k_2 \),令\( a_i := 0 \),易证\( \pm \sum_{i = -\infty}^{k_2} a_i \times 10^i = \pm \sum_{i = -\infty}^{k + m} a_i \times 10^i \),这样\( \pm a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots = \pm \sum_{i = -\infty}^{k_2} a_i \times 10^i \)才是一个合法的\( x \)的十进制表示,针对该十进制表示,易得\( \forall i < m, a_i = 0 \),又\( a_{k + m} > 0 \),于是可得\( \exists i_0 \)满足\( m \leq i_0 \leq k + m, a_{i_0} > 0 \)且\( \forall i < i_0, a_i = 0 \) (\( a_{i_0} \)即最右侧的非\( 0 \)数字)。构造一个十进制表示\( \pm b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots = \pm \sum_{i = -\infty}^{k_2} b_i \times 10^i, \forall i \leq k_2 \),如果\( i_0 < i \leq k_2 \),则令\( b_i := a_i \),如果\( i = i_0 \),则令\( b_i := a_{i_0} - 1 \),如果\( i < i_0 \),则令\( b_i := 9 \),即右侧的最后一位非\( 0 \)数字\( - 1 \),然后剩余的数字全部改成\( 9 \),但是注意到,如果原始的最后一位非\( 0 \)数字\( a_{i_0} \)刚好也是最左边的非\( 0 \)数字,即\( i_0 = k_2 \),同时还满足\( a_{i_0} = 1 \)的话,则\( b_{k_2} = b_{i_0} = a_{i_0} - 1 = 0 \),这违反了十进制表示法的约定,得修正下,我们定义新的级数下标上限\( k_3 \),如果\( b_{k_2} = 0 \),则令\( k_3 := k_2 - 1 \)(注:此时\( b_{k_3} = 9 \)),否则令\( k_3 := k_2 \),易得\( \pm \sum_{i = -\infty}^{k_3} b_i \times 10^i = \pm \sum_{i = -\infty}^{k_2} b_i \times 10^i \)。我们得证明下\( \pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)是\( x \)的十进制表示, \( \pm \sum_{i = -\infty}^{k_3} b_i \times 10^i = \pm ((\sum_{i = i_0 + 1}^{k_3} a_i \times 10^i) + ((a_{i_0} - 1) \times 10^{i_0}) + (\sum_{i = -\infty}^{i_0 - 1} 9 \times 10^i)) \),易证\( ((a_{i_0} - 1) \times 10^{i_0}) + (\sum_{i = -\infty}^{i_0 - 1} 9 \times 10^i) = a_{i_0} \times 10^{i_0} = \sum_{i = -\infty}^{i_0} a_i \times 10^i \) (注:\( \forall i < i_0, a_i = 0 \)),因此\( \pm \sum_{i = -\infty}^{k_3} b_i \times 10^i = \pm ((\sum_{i = i_0 + 1}^{k_3} a_i \times 10^i) + \sum_{i = -\infty}^{i_0} a_i \times 10^i) = \pm \sum_{i = -\infty}^{k_3} a_i \times 10^i = x \),即\( \pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)是\( x \)的十进制表示。综上,\( x \)有两个十进制表示且明显这两个十进制表示是不同的(比如第\( i_0 \)位的数字不同)。还得证明\( x \)至多只有两个不同的十进制表示: \( \forall x \)的十进制表示\( \pm c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \),如果\( k_4 > k_2 \),则\( |x| = |\pm c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots| = c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots = \sum_{i = -\infty}^{k_4} c_i \times 10^i \geq \sum_{i = i_0}^{k_4} c_i \times 10^i > \sum_{i = i_0}^{k_2} a_i \times 10^i = \sum_{i = -\infty}^{k_2} a_i \times 10^i \text{(注:} \forall i < i_0, a_i = 0 \text{)} = a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots = |\pm a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots| = |x| \),即\( |x| > |x| \),矛盾,因此\( k_4 \leq k_2 \)。假设\( k_4 < k_3 \),则\( |x| = |\pm c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots| = c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots = \sum_{i = -\infty}^{k_4} c_i \times 10^i \leq \sum_{i = i_0}^{k_3} b_i \times 10^i < (\sum_{i = i_0}^{k_3} b_i \times 10^i) + 10^{i_0} = (\sum_{i = i_0}^{k_3} b_i \times 10^i) + (\sum_{i = -\infty}^{i_0 - 1} b_i \times 10^i) \text{(注:} \forall i < i_0, b_i = 9 \text{)} = \sum_{i = -\infty}^{k_3} b_i \times 10^i = b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots = |\pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots| = |x| \),即\( |x| < |x| \),矛盾,因此\( k_4 \geq k_3 \),加上前面得到的\( k_4 \leq k_2 \)以及\( k_3 = k_2 \)或\( k_3 = k_2 - 1 \),可得\( k_4 = k_2 \)或\( k_4 = k_3 \)。 \( \pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)是从\( \pm a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots \)构造出来的,因此两者正负号相同,而由于\( x \neq 0 \),因此\( \pm c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \)的正负号也必须 \( \pm a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots, \pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \) 相同,否则级数的和的正负号会不同,这和它们是同一个实数的十进制表示矛盾,故接下来,我们将不怎么讨论正负号的问题。接下来分情况讨论:

  1. 如果\( \exists i_1 \leq k_4, \forall i < i_1, c_i = 0 \),我们想证明十进制表示\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \) 和十进制表示\( a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots \)相同,假设十进制表示\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \) 和十进制表示\( a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots \)不同 (1) ,令\( i_3 := \min(i_0, i_1) \),则此时\( |x| \times 10^{-i_3} = (c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots) \times 10^{-i_3} = (\sum_{i = -\infty}^{k_4} c_i \times 10^i) \times 10^{-i_3} = (\sum_{i = i_1}^{k_4} c_i \times 10^i) \times 10^{-i_3} = \sum_{i = i_1}^{k_4} c_i \times 10^{i - i_3} = \sum_{i = i_1 - i_3}^{k_4 - i_3} c_{i + i_3} \times 10^i \),又\( i_3 \leq i_1 \),因此下标下限\( i_1 - i_3 \geq 0 \),构造不带小数点的十进制表示\( d_{k_4 - i_3} \dots d_0, \forall 0 \leq i \leq k_4 - i_3 \),如果\( i \geq i_1 - i_3 \),则令\( d_i := c_{i + i_3} \),否则,令\( d_i := 0 \),可得\( d_{k_4 - i_3} \dots d_0 \)为正整数\( |x| \times 10^{-i_3} \)不带小数点的十进制表示,又\( |x| \times 10^{-i_3} = (a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots) \times 10^{-i_3} = (\sum_{i = -\infty}^{k_2} a_i \times 10^i) \times 10^{-i_3} = (\sum_{i = i_0}^{k_2} a_i \times 10^i) \times 10^{-i_3} = \sum_{i = i_0}^{k_2} a_i \times 10^{i - i_3} = \sum_{i = i_0 - i_3}^{k_2 - i_3} a_{i + i_3} \times 10^i \),又\( i_3 \leq i_0 \),因此下标下限\( i_0 - i_3 \geq 0 \),构造不带小数点的十进制表示\( e_{k_2 - i_3} \dots e_0, \forall 0 \leq i \leq k_2 - i_3 \),如果\( i \geq i_0 - i_3 \),则令\( d_i := a_{i + i_3} \),否则,令\( d_i := 0 \),可得\( e_{k_2 - i_3} \dots e_0 \)也为正整数\( |x| \times 10^{-i_3} \)不带小数点的十进制表示,但由(1)以及我们仅仅只是乘以\( 10^{-i_3} \)左移或者右移两个十进制表示\( |i_3| \)位而已,可得位移后的两个十进制表示\( d_{k_4 - i_3} \dots d_0, e_{k_2 - i_3} \dots e_0 \)也是不同的,即正整数\( |x| \times 10^{-i_3} \)有两个不同的不带小数点的十进制表示,然而这和定理B.1.4矛盾,因此假设不成立,有\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \) 和十进制表示\( a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots \)相同。
  2. 如果上一种情况不满足,即\( \forall N \leq k_4, \exists i < N, c_i > 0 \) (2) (注:由于\( c_i \)是非负数字,因此\( c_i \neq 0 \)相当于\( c_i > 0 \)),我们想证明十进制表示\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots \) 和十进制表示\( b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)相同,我们知道\( k_4 = k_2 \)或\( k_4 = k_3 \),这里我们首先证明\( k_4 = k_3 \),特别注意,此时仍然可能\( k_4 = k_2 \),这是因为\( k_2 = k_3 \)或\( k_2 = k_3 + 1 \) (3) ,假设\( k_4 > k_3 \),则\( |x| = c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots = \sum_{i = -\infty}^{k_4} c_i \times 10^i = ((c_{k_4} - 1) \times 10^{k_4}) + 10^{k_4} + \sum_{i = -\infty}^{k_4 - 1} c_i \times 10^i \geq ((c_{k_4} - 1) \times 10^{k_4}) + (\sum_{i = -\infty}^{k_3} b_i \times 10^i) + \sum_{i = -\infty}^{k_4 - 1} c_i \times 10^i > \sum_{i = -\infty}^{k_3} b_i \times 10^i \text{(注:这里严格大于是由(2)得到的,因为由(2)可得,} \sum_{i = -\infty}^{i_0 - 1} c_i \times 10^i > 0 \text{)} = b_{k_3} \dots b_0 . b_{-1} a_{-2} \dots = |x| \),即\( |x| > |x| \)矛盾,因此假设不成立,有\( k_4 \leq k_3 \),加上(3),可得\( k_4 = k_3 \)。现在,我们使用强归纳法的反向归纳的版本来证明(或者可以把下标取反,但那样可读性差),归纳假设\( \forall i_c < j \leq k_4, \forall j \leq i \leq k_4, c_i = b_i \),我们需要证明\( \forall i_c \leq i \leq k_4, c_i = b_i \),根据归纳假设, \( \forall i_c + 1 \leq i \leq k_4, c_i = b_i \),因此我们只需要证明\( c_{i_c} = b_{i_c} \)即可,除此之外,还有\( \sum_{i = i_c + 1}^{k_4} c_i \times 10^i = \sum_{i = i_c + 1}^{k_4} b_i \times 10^i \),假设\( c_{i_c} > b_{i_c} \),则\( |x| = c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots = \sum_{i = -\infty}^{k_4} c_i \times 10^i = (\sum_{i = i_c + 1}^{k_4} c_i \times 10^i) + ((c_{i_c} - 1) \times 10^{i_c}) + 10^{i_c} + (\sum_{i = -\infty}^{i_c - 1} c_i \times 10^i) \geq (\sum_{i = i_c + 1}^{k_4} c_i \times 10^i) + ((c_{i_c} - 1) \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} c_i \times 10^i) \geq (\sum_{i = i_c + 1}^{k_4} c_i \times 10^i) + (b_{i_c} \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} c_i \times 10^i) = (\sum_{i = i_c + 1}^{k_4} b_i \times 10^i) + (b_{i_c} \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} c_i \times 10^i) = (\sum_{i = -\infty}^{k_4} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} c_i \times 10^i) > \sum_{i = -\infty}^{k_4} b_i \times 10^i = \sum_{i = -\infty}^{k_3} b_i \times 10^i = b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots = |x| \),即\( |x| > |x| \),矛盾,因此假设不成立,有\( c_{i_c} \leq b_{i_c} \),假设\( c_{i_c} < b_{i_c} \),则同上,将\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots, b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)对应交换,易得\( |x| < |x| \),矛盾,因此假设不成立,至此可得,\( c_{i_c} = b_{i_c} \),加上根据归纳假设得到的\( \forall i_c + 1 \leq i \leq k_4, c_i = b_i \),有\( \forall i_c \leq i \leq k_4, c_i = b_i \),归纳完毕,加上\( k_4 = k_3 \),可得\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots, b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots = |x| \)为相同的十进制表示。

综上,如果\( \exists i_1 \leq k_4, \forall i < i_1, c_i = 0 \),则\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots, a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots \)为相同的十进制表示,否则,\( c_{k_4} \dots c_0 . c_{-1} c_{-2} \dots, b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)为相同的十进制表示,综合可得, \( x \)有且仅有两个十进制表示\( \pm a_{k_2} \dots a_0 . a_{-1} a_{-2} \dots, \pm b_{k_3} \dots b_0 . b_{-1} b_{-2} \dots \)。

如果\( x = 0 \):我们知道它至少有一个十进制表示\( 0.000 \dots \),得排除它有一个以上的十进制表示:假设它还有另外一个不同的十进制表示 \( \pm a_k \dots a_0 . a_{-1} a_{-2} \dots = \pm \sum_{i = -\infty}^{k} a_i \times 10^i \),由于该十进制表示和 \( 0.000 \dots \)不同,于是\( \exists i_0 \leq k, a_{i_0} > 0 \),进而\( |\pm \sum_{i = -\infty}^{k} a_i \times 10^i| \geq a_{i_0} \times 10^{i_0} > 0 \),这和\( \pm a_k \dots a_0 . a_{-1} a_{-2} \dots \)是\( 0 \)的十进制表示矛盾,于是假设不成立,有\( 0 \)有且仅有一个十进制表示。

如果\( x \neq 0 \)且不是有尽十进制:根据定理B.2.2,我们知道它至少有一个十进制表示,得排除它有一个以上的十进制表示:\( \forall x \)的十进制表示 \( \pm a_k \dots a_0 . a_{-1} a_{-2} \dots = \pm \sum_{i = -\infty}^{k} a_i \times 10^i \),我们证明该十进制表示的尾部不能全是\( 0 \),即\( \forall N \leq k, \exists i < N, a_i > 0 \):假设\( \exists N > 0, \forall i < N, a_i = 0 \),又\( x \neq 0 \),可得\( a_k > 0 \),于是\( \exists i_0 \)满足\( N \leq i_0 \leq k, a_{i_0} > 0, \forall i < i_0, a_{i_0} = 0 \),令\( n := (\pm \sum_{i = i_0}^{k} a_i \times 10^i) \times 10^{-i_0}, m := -i_0 \),可得\( n, m \)为整数且\( x = n / 10^{-m} \),这意味着\( x \)是有尽十进制,矛盾,于是假设不成立,有\( \forall N \leq k, \exists i < N, a_i > 0 \) (4) 。 \( \forall x \)的两个十进制表示\( \pm a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots, \pm b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots \),我们想证明这两个十进制表示相同,首先,这两个十进制表示的正负号必须相同,否则级数的和的正负号会不同,从而和这两个十进制表示是同一个实数的十进制表示矛盾,故之后,我们将不怎么讨论正负号的问题。接下来,我们先证明\( k_1 = k_2 \),假设\( k_1 > k_2 \),则\( |x| = a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots = \sum_{i = -\infty}^{k_1} a_i \times 10^i = (a_{k_1} \times 10^{k_1}) + (\sum_{i = -\infty}^{k_1 - 1} a_i \times 10^i) \geq (\sum_{i = -\infty}^{k_2} b_i \times 10^i) + (\sum_{i = -\infty}^{k_1 - 1} a_i \times 10^i) > \sum_{i = -\infty}^{k_2} b_i \times 10^i \text{(注:这里严格大于是由(4)得到的,因为由(4)可得,} \sum_{i = -\infty}^{k_1 - 1} a_i \times 10^i > 0 \text{)} = b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots = |x| \),即\( |x| > |x| \),矛盾,因此假设不成立,有\( k_1 \leq k_2 \),假设\( k_1 < k_2 \),则同上,将\( a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots, b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots \)对应交换,易得\( |x| < |x| \),矛盾,因此假设不成立,至此可得,\( k_1 = k_2 \)。现在,我们使用强归纳法的反向归纳的版本来证明,归纳假设\( \forall i_c < j \leq k_1, \forall j \leq i \leq k_1, b_i = a_i \),我们需要证明\( \forall i_c \leq i \leq k_1, b_i = a_i \),根据归纳假设, \( \forall i_c + 1 \leq i \leq k_1, b_i = a_i \),因此我们只需要证明\( b_{i_c} = a_{i_c} \)即可,除此之外,还有\( \sum_{i = i_c + 1}^{k_1} b_i \times 10^i = \sum_{i = i_c + 1}^{k_1} a_i \times 10^i \),假设\( b_{i_c} > a_{i_c} \),则\( |x| = a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots = \sum_{i = -\infty}^{k_1} a_i \times 10^i = (\sum_{i = i_c + 1}^{k_1} a_i \times 10^i) + ((a_{i_c} - 1) \times 10^{i_c}) + 10^{i_c} + (\sum_{i = -\infty}^{i_c - 1} a_i \times 10^i) \geq (\sum_{i = i_c + 1}^{k_1} a_i \times 10^i) + ((a_{i_c} - 1) \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} a_i \times 10^i) \geq (\sum_{i = i_c + 1}^{k_1} a_i \times 10^i) + (b_{i_c} \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} a_i \times 10^i) = (\sum_{i = i_c + 1}^{k_1} b_i \times 10^i) + (b_{i_c} \times 10^{i_c}) + (\sum_{i = -\infty}^{i_c - 1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} a_i \times 10^i) = (\sum_{i = -\infty}^{k_1} b_i \times 10^i) + (\sum_{i = -\infty}^{i_c - 1} a_i \times 10^i) > \sum_{i = -\infty}^{k_1} b_i \times 10^i = \sum_{i = -\infty}^{k_2} b_i \times 10^i = b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots = |x| \),即\( |x| > |x| \),矛盾,因此假设不成立,有\( a_{i_c} \leq b_{i_c} \),假设\( a_{i_c} < b_{i_c} \),则同上,将\( a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots, b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots \)对应交换,易得\( |x| < |x| \),矛盾,因此假设不成立,至此可得,\( a_{i_c} = b_{i_c} \),加上归纳假设得到的\( \forall i_c + 1 \leq i \leq k_1, b_i = a_i \),有\( \forall i_c \leq i \leq k_1, b_i = a_i \),归纳完毕,加上\( k_1 = k_2 \),可得\( a_{k_1} \dots a_0 . a_{-1} a_{-2} \dots, b_{k_2} \dots b_0 . b_{-1} b_{-2} \dots \) 是相同的十进制表示。综上,\( x \)有且仅有一个十进制表示。

证毕。

练习B.2.4

题目:

Rewrite the proof of Corollary 8.3.4 using the decimal system.

Corollary 8.3.4的内容:

\( \mathbf{R} \) is uncountable.

注:

使用康托的对角线证明法。

证明:

假设\( \mathbf{R} \)可数,则\( \exists \)双射函数\( f: \mathbf{N} \to \mathbf{R} \),构造\( y := \sum_{i = -\infty}^{0} a_i \times 10^i, \forall i \leq 0 \),令\( x_i := f(-i) \),令\( \sum_{j = -\infty}^{k_{i1}} b_j \times 10^j, \sum_{j = -\infty}^{k_{i2}} c_j \times 10^j \)为\( x_i \)对应的两个十进制表示,根据练习B.2.3,\( x_i \)有一个或者两个十进制表示,如果它只有一个十进制表示 \( \sum_{j = -\infty}^{k_{i1}} b_j \times 10^j \),则我们令\( \sum_{j = -\infty}^{k_{i2}} c_j \times 10^j := \sum_{j = -\infty}^{k_{i1}} b_j \times 10^j \),可得\( \sum_{j = -\infty}^{k_{i1}} b_j \times 10^j, \sum_{j = -\infty}^{k_{i2}} c_j \times 10^j \) 是\( x_i \)所有的十进制表示,如果\( i > k_{i1} \),则我们约定\( x_i \)第1个十进制表示的第\( i \)位数字\( b_i := 0 \),如果\( i > k_{i2} \),则我们约定\( x_i \)第2个十进制表示的第\( i \)位数字\( c_i := 0 \),令\( a_i \)为不同于\( b_i \)且不同于\( c_i \)的数字且\( > 0 \)的数字(选\( > 0 \)是为了确保\( a_0 > 0 \)),由于\( b_i, c_i \)最多只占两个数字,再排除掉\( 0 \),至少还剩\( 10 - 2 - 1 = 7 \)个数字可以选,\( a_i \)是绝对可以选到和\( b_i, c_i \)不同且\( > 0 \)的数字的。根据练习B.2.1,\( y \)是一个绝对收敛级数,因此\( y \)是一个实数,但\( \forall i \in \mathbf{N}, y \)的第\( -i \)位数字均和\( f(i) = x_{-i} \)的 所有 十进制表示的第\( i \)位数字不同,因此\( y \neq f(i) \),综上,有\( \forall i \in \mathbf{N}, f(i) \neq y \),然而\( y \in \mathbf{R} \),这和\( f \)满射矛盾,因此假设不成立,有\( \mathbf{R} \)不可数。

证毕。