1 − 2 + 3 − 4 + …:修订间差异

维基百科,自由的百科全书
删除的内容 添加的内容
无编辑摘要
第1行: 第1行:
{{Translating|tfrom=[[:en:1 − 2 + 3 − 4 + · · ·]]|tpercent=86}}
{{Translating|tfrom=[[:en:1 − 2 + 3 − 4 + · · ·]]|tpercent=95}}
[[Image:Pm1234 Ground.png|thumb|1 − 2 + 3 − 4 + …的前几千项相加结果示意图]]
[[Image:Pm1234 Ground.png|thumb|1 − 2 + 3 − 4 + …的前几千项相加结果示意图]]


第12行: 第12行:
:<math>1-2+3-4+\cdots=\frac{1}{4}.</math>
:<math>1-2+3-4+\cdots=\frac{1}{4}.</math>


该等式的[[严密]]解释在很久以后才出现。[[1890年]]初,[[:en:Ernesto Cesàro|恩纳斯托·切萨罗]]、[[埃米尔·博雷尔]]与其他一些数学家研究出了[[:en:well-defined|定义明确]](well-defined)的方法,来为有穷极限确定广义级数的和——其中包含了欧拉所寻找的新解释。大部分这些可求和法最终可简单地确定 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的“和”为<sup>1</sup>⁄<sub>4</sub>。[[:en:Cesàro summation|切萨罗求和]](Cesàro summation)是少数几种不能计算出 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 之和的方法。因此此级数是需要某个略强的方法——譬如[[:en:Divergent series#Abelian means|阿贝耳求和]](Abel summation)——的例子。
该等式的[[严密]]解释在很久以后才出现。[[1890年]]初,[[恩纳斯托·切萨罗]]、[[埃米尔·博雷尔]]与其他一些数学家研究出了[[定义明确]]([[:en:well-defined|well-defined]]的方法,来为有穷极限确定广义级数的和——其中包含了欧拉所寻找的新解释。大部分这些可求和法最终可简单地确定{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的“和”为<sup>1</sup>⁄<sub>4</sub>。[[切萨罗求和]]([[:en:Cesàro summation|Cesàro summation]]是少数几种不能计算出{{nowrap|1 − 2 + 3 − 4 + &hellip;}}之和的方法。因此此级数是需要某个略强的方法——譬如[[阿贝耳求和]]([[:en:Divergent series#Abelian means|Abel summation]])——的例子。


级数 {{nowrap|1, −1, 2, −2, &hellip;}} 与[[:en:Grandi's series|格兰第级数]] {{nowrap|1 − 1 + 1 − 1 + &hellip;}} 有较近的联系。欧拉将此二级数视作{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}}对任意''n''的特殊情况;这是他在[[:en:Basel problem|巴塞尔问题]](Basel problem)上所作工作的一个延伸研究方向,并导向了[[:en:functional equation|泛函方程]]——即我们现在所知的[[:en:Dirichlet eta function|狄利克η函数]]与[[黎曼ζ函数]]。
级数{{nowrap|1, −1, 2, −2, &hellip;}}与[[格兰第级数]]([[:en:Grandi's series|Grandi's series]]{{nowrap|1 − 1 + 1 − 1 + &hellip;}}有较近的联系。欧拉将此二级数视作{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}}对任意''n''的特殊情况;这是他在[[巴塞尔问题]]([[:en:Basel problem|Basel problem]]上所作工作的一个延伸研究方向,并导向了[[泛函方程]]——即我们现在所知的[[狄利克η函数]]与[[黎曼ζ函数]]。


== 发散 ==
== 发散 ==
级数项(1, −2, 3, −4, &hellip;)没有趋近于[[0]];因此可通过[[:en:term test|项测试]](term test)来确定 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 发散。作为后文的参考,此方法也常被用于从基础级上预见发散。从定义可知,无穷级数的收敛或发散是由其部分和的[[收敛数列|收敛或发散]]确定的, {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的部分和为:<ref>Hardy p.8</ref>
级数项(1, −2, 3, −4, &hellip;)没有趋近于[[0]];因此可通过[[项测试]]([[:en:term test|term test]]来确定{{nowrap|1 − 2 + 3 − 4 + &hellip;}}发散。作为后文的参考,此方法也常被用于从基础级上预见发散。从定义可知,无穷级数的收敛或发散是由其部分和的[[收敛数列|收敛或发散]]确定的,{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的部分和为:<ref>Hardy p.8</ref>
:1 = '''1''',
:1 = '''1''',
:1 − 2 = '''−1''',
:1 − 2 = '''−1''',
第25行: 第25行:
:1 − 2 + 3 − 4 + 5 − 6 = '''−3''',
:1 − 2 + 3 − 4 + 5 − 6 = '''−3''',
:&hellip;
:&hellip;
此序列的一个显著的特点是每个整数都出现了一次——如果将空部分和计入还包括0——因此其[[可数集]]为整数集<math>\mathbb{Z}</math>。<ref>Beals p.23</ref>很明显的,不可能让变化的结果辐合到一个确定的数,因此 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 发散。
此序列的一个显著的特点是每个整数都出现了一次——如果将空部分和计入还包括0——因此其[[可数集]]为整数集<math>\mathbb{Z}</math>。<ref>Beals p.23</ref>很明显的,不可能让变化的结果辐合到一个确定的数,因此{{nowrap|1 − 2 + 3 − 4 + &hellip;}}发散。


== 求和的[[启发法|启发]] ==
== 求和的[[启发法|启发]] ==
=== 稳定性与线性 ===
=== 稳定性与线性 ===
由于各项 1, −2, 3, −4, 5, −6, &hellip; 以一种简单模式排列,级数 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 可表示为它自己的变换形式(概略的说法),而因此所得的[[方程求解|方程解]]取得一数值。暂时假设写作 {{nowrap|1=''s'' = 1 − 2 + 3 − 4 + &hellip;}} 有意义——其中的''s''为常数,然后再处理{{nowrap|1=''s'' = <sup>1</sup>⁄<sub>4</sub>:}}的问题:
由于各项 1, −2, 3, −4, 5, −6, &hellip; 以一种简单模式排列,级数{{nowrap|1 − 2 + 3 − 4 + &hellip;}}可表示为它自己的变换形式(概略的说法),而因此所得的[[方程求解|方程解]]取得一数值。暂时假设写作{{nowrap|1=''s'' = 1 − 2 + 3 − 4 + &hellip;}}有意义——其中的''s''为常数,然后再处理{{nowrap|1=''s'' =<sup>1</sup>⁄<sub>4</sub>:}}的问题:


[[Image:Pm1234 linearity.png|thumb|right|复制4份 1 − 2 + 3 − 4 + &hellip;,仅使用移动与项项相加,结果为1。左右两边的两个 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 副本相互抵消,并得出 {{nowrap|1 − 1 + 1 − 1 + &hellip;}}。]]
[[Image:Pm1234 linearity.png|thumb|right|复制4份 1 − 2 + 3 − 4 + &hellip;,仅使用移动与项项相加,结果为1。左右两边的两个{{nowrap|1 − 2 + 3 − 4 + &hellip;}}副本相互抵消,并得出{{nowrap|1 − 1 + 1 − 1 + &hellip;}}。]]
:{|border="0" cellpadding="0" cellspacing="12" style="font-size:9pt;"
:{|border="0" cellpadding="0" cellspacing="12" style="font-size:9pt;"
|-
|-
第71行: 第71行:
|}
|}


因此,{{nowrap|1=''s'' = <sup>1</sup>⁄<sub>4</sub>}}.<ref>Hardy (p.6) presents this derivation in conjunction with evaluation of [[Grandi's series]] {{nowrap|1 − 1 + 1 − 1 + &hellip;}}.</ref>,如右图所示。
因此,{{nowrap|1=''s'' =<sup>1</sup>⁄<sub>4</sub>}}.<ref>Hardy (p.6) 结合[[格兰第级数]]{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的计算提出了此推导过程。</ref>,如右图所示。


尽管 1 − 2 + 3 − 4 + &hellip; 没有通常意义的和,等式 {{nowrap|1=''s'' = 1 − 2 + 3 − 4 + &hellip; = <sup>1</sup>⁄<sub>4</sub>}} 却可被赋予另外一种意义。离散级数之“和”的一种[[广义化|普遍]]定义被称为一种[[:en:summation method|求和法]](summation method)或[[:en:summability method|可和法]](summability method)——对所有可能级数的一些子集求和。有许多种不同的方法(部分将在[[#特殊方法|下文]]中出现),这些方法有着与常数求和所共有的特性。以上的处理实际上都可由下述所证明:给出任意的[[:en:Divergent series#Properties of summation methods|线形且稳定]]的可和法,并计算级数 {{nowrap|1=1 − 2 + 3 − 4 + &hellip;}} ,结果为 <sup>1</sup>⁄<sub>4</sub>。此外,由于:
尽管1 − 2 + 3 − 4 + &hellip;没有通常意义的和,等式{{nowrap|1=''s'' = 1 − 2 + 3 − 4 + &hellip; =<sup>1</sup>⁄<sub>4</sub>}}却可被赋予另外一种意义。离散级数之“和”的一种[[广义化|普遍]]定义被称为一种[[:en:summation method|求和法]](summation method)或[[:en:summability method|可和法]](summability method)——对所有可能级数的一些子集求和。有许多种不同的方法(部分将在[[#特殊方法|下文]]中出现),这些方法有着与常数求和所共有的特性。以上的处理实际上都可由下述所证明:给出任意的[[:en:Divergent series#Properties of summation methods|线形且稳定]]的可和法,并计算级数{{nowrap|1=1 − 2 + 3 − 4 + &hellip;}},结果为<sup>1</sup>⁄<sub>4</sub>。此外,由于:


:{|border="0" cellpadding="0" cellspacing="12" style="font-size:9pt;"
:{|border="0" cellpadding="0" cellspacing="12" style="font-size:9pt;"
第98行: 第98行:
| colspan=6 | = 1 - 1 + 1 - 1 &hellip;,
| colspan=6 | = 1 - 1 + 1 - 1 &hellip;,
|}
|}
故此方法也一定能对[[格兰第级数]]求和,并得结果为{{nowrap|1=1 − 1 + 1 − 1 + &hellip; = <sup>1</sup>⁄<sub>2</sub>}}。
故此方法也一定能对[[格兰第级数]]求和,并得结果为{{nowrap|1=1 − 1 + 1 − 1 + &hellip; =<sup>1</sup>⁄<sub>2</sub>}}。


以上为 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的一般和限制了可能的值,但依然不能首先显现出哪种方法能或不能计算出此级数来。事实上,一些线形且稳定的可和法,譬如常数求和,是无法计算出 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的。看待此级数的不同眼光有助于确定哪些方法可计算出 <sup>1</sup>⁄<sub>4</sub>:表现为[[积]]的形式。
以上为{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的一般和限制了可能的值,但依然不能首先显现出哪种方法能或不能计算出此级数来。事实上,一些线形且稳定的可和法,譬如常数求和,是无法计算出{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的。看待此级数的不同眼光有助于确定哪些方法可计算出<sup>1</sup>⁄<sub>4</sub>:表现为[[积]]的形式。


=== 柯西乘积 ===
=== 柯西乘积 ===
[[1891年]],, [[恩纳斯托·切萨罗]]([[:en:Ernesto Cesàro|Ernesto Cesàro]])表示了将分散级数严密地带入[[微积分学]]的愿望,并指出:“已可写出{{nowrap|1=(1 − 1 + 1 − 1 + &hellip;)<sup>2</sup> = 1 − 2 + 3 − 4 + &hellip;}}并断定两边均等于<sup>1</sup>⁄<sub>4</sub>。”<ref>"One already writes {{nowrap|1=(1 − 1 + 1 − 1 + &hellip;)<sup>2</sup> = 1 − 2 + 3 − 4 + &hellip;}} and asserts that both the sides are equal to <sup>1</sup>⁄<sub>4</sub>.", Ferraro, p.130.</ref>对切萨罗而言,这个等式是他前几年发表的一个定理的应用,该定理也许是历史上可求和的发散级数的第一个定理。关于此求和法的详细内容请见[[#切萨罗与赫尔德|下文]];其中心思想是 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} {{nowrap|1 − 1 + 1 − 1 + &hellip;}} {{nowrap|1 − 1 + 1 − 1 + &hellip;}} 的[[柯西乘积]]([[:en:Cauchy product|Cauchy product]])。
[[1891年]],, [[恩纳斯托·切萨罗]]([[:en:Ernesto Cesàro|Ernesto Cesàro]])表示了将分散级数严密地带入[[微积分学]]的愿望,并指出:“已可写出{{nowrap|1=(1 − 1 + 1 − 1 + &hellip;)<sup>2</sup> = 1 − 2 + 3 − 4 + &hellip;}}并断定两边均等于<sup>1</sup>⁄<sub>4</sub>。”<ref>"One already writes{{nowrap|1=(1 − 1 + 1 − 1 + &hellip;)<sup>2</sup> = 1 − 2 + 3 − 4 + &hellip;}}and asserts that both the sides are equal to<sup>1</sup>⁄<sub>4</sub>.", Ferraro, p.130.</ref>对切萨罗而言,这个等式是他前几年发表的一个定理的应用,该定理也许是历史上可求和的发散级数的第一个定理。关于此求和法的详细内容请见[[#切萨罗与赫尔德|下文]];其中心思想是{{nowrap|1 − 2 + 3 − 4 + &hellip;}}是{{nowrap|1 − 1 + 1 − 1 + &hellip;}}对{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的[[柯西乘积]]([[:en:Cauchy product|Cauchy product]])。


[[Image:Pm1234 Cauchy.svg|thumb|left|1 − 2 + 3 − 4 + &hellip; 以 1 − 1 + 1 − 1 + &hellip; 的二重柯西乘积出现]]
[[Image:Pm1234 Cauchy.svg|thumb|left|1 − 2 + 3 − 4 + &hellip; 以 1 − 1 + 1 − 1 + &hellip; 的二重柯西乘积出现]]
两个无穷级数的柯西乘积可被确定,即使在他们都发散的时候。在 Σ''a''<sub>''n''</sub> = Σ''b''<sub>''n''</sub> = Σ(−1)<sup>''n''</sup> 的情况下,柯西乘积的项可由有穷对角线求和的方式给出:
两个无穷级数的柯西乘积可被确定,即使在他们都发散的时候。在 Σ''a''<sub>''n''</sub>= Σ''b''<sub>''n''</sub>= Σ(−1)<sup>''n''</sup> 的情况下,柯西乘积的项可由有穷对角线求和的方式给出:


:<math>\begin{array}{rcl}
:<math>\begin{array}{rcl}
第116行: 第116行:
:<math>\sum_{n=0}^\infty(-1)^n(n+1) = 1-2+3-4+\cdots.</math>
:<math>\sum_{n=0}^\infty(-1)^n(n+1) = 1-2+3-4+\cdots.</math>


这样一种考虑到两个级数的柯西乘积与{{nowrap|1=1 − 1 + 1 − 1 + &hellip; = <sup>1</sup>⁄<sub>2</sub>}}结果的求和法,也能够求出{{nowrap|1=1 − 2 + 3 − 4 + &hellip; = <sup>1</sup>⁄<sub>4</sub>}}。由前部分的结果可知,当方法是线形、稳定并考虑到柯西乘积的时候,{{nowrap|1 − 1 + 1 − 1 + &hellip;}} {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的可求和之间是等价的。
这样一种考虑到两个级数的柯西乘积与{{nowrap|1=1 − 1 + 1 − 1 + &hellip; =<sup>1</sup>⁄<sub>2</sub>}}结果的求和法,也能够求出{{nowrap|1=1 − 2 + 3 − 4 + &hellip; =<sup>1</sup>⁄<sub>4</sub>}}。由前部分的结果可知,当方法是线形、稳定并考虑到柯西乘积的时候,{{nowrap|1 − 1 + 1 − 1 + &hellip;}}与{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的可求和之间是等价的。


切萨罗的定理是一个深奥的例子。级数 {{nowrap|1=1 − 1 + 1 − 1 + &hellip;}} 在最弱的意义上是切萨罗可求和,称作 {{nowrap|(C, 1)-可求和}},然而 {{nowrap|1=1 − 2 + 3 − 4 + &hellip;}} 则需要切萨罗的定理的一个更强的形式<ref>Hardy, p.3; Weidlich, pp.52–55.</ref>,表示为 {{nowrap|(C, 2)-可求和}}。由于切萨罗的定理的所有形式均为线形且稳定的,所得的值正是此前计算所得的。
切萨罗的定理是一个深奥的例子。级数{{nowrap|1=1 − 1 + 1 − 1 + &hellip;}}在最弱的意义上是切萨罗可求和,称作{{nowrap|(C, 1)-可求和}},然而{{nowrap|1=1 − 2 + 3 − 4 + &hellip;}}则需要切萨罗的定理的一个更强的形式<ref>Hardy, p.3; Weidlich, pp.52–55.</ref>,表示为{{nowrap|(C, 2)-可求和}}。由于切萨罗的定理的所有形式均为线形且稳定的,所得的值正是此前计算所得的。


== 特殊方法 ==
== 特殊方法 ==
=== 切萨罗与赫尔德 ===
=== 切萨罗与赫尔德 ===
[[Image:Pm1234 means.svg|thumb|right|关于 <sup>1</sup>⁄<sub>4</sub> (H, 2) 和的数据]]
[[Image:Pm1234 means.svg|thumb|right|关于<sup>1</sup>⁄<sub>4</sub>的(H, 2)和的数据]]


要找到1 − 2 + 3 − 4 + &hellip的 (C, 1) [[切萨罗求和|切萨罗和]],如果它存在,就需要计算该级数部分和的[[算术平均值]]。
要找到1 − 2 + 3 − 4 + &hellip的(C, 1)[[切萨罗求和|切萨罗和]],如果它存在,就需要计算该级数部分和的[[算术平均值]]。
部分和为:
部分和为:


第131行: 第131行:
这些部分和的算术平均值为:
这些部分和的算术平均值为:


:1, 0, <sup>2</sup>⁄<sub>3</sub>, 0, <sup>3</sup>⁄<sub>5</sub>, 0, <sup>4</sup>⁄<sub>7</sub>, &hellip;.
:1, 0,<sup>2</sup>⁄<sub>3</sub>, 0,<sup>3</sup>⁄<sub>5</sub>, 0,<sup>4</sup>⁄<sub>7</sub>, &hellip;.


此平均值序列没有辐合,因此 1 − 2 + 3 − 4 + &hellip; 不是切萨罗可求和。
此平均值序列没有辐合,因此 1 − 2 + 3 − 4 + &hellip; 不是切萨罗可求和。


切萨罗求和有两种有名的广义化:让这些在概念上更简单的是 (H, ''n'') 法的序列,其中''n''为[[自然数]]。(H, 1) 和为切萨罗求和,更高的方法则重复平均值的计算。在上文中,偶数平均值趋近于 <sup>1</sup>⁄<sub>2</sub>,奇数平均值则全部等于0,所以平均值''的''平均值趋近于 0 与 <sup>1</sup>⁄<sub>2</sub> 的平均数,即 <sup>1</sup>⁄<sub>4</sub>。<ref>Hardy, p.9. For the full details of the calculation, see Weidlich, pp.17–18.</ref>因此, {{nowrap|1 − 2 + 3 − 4 + &hellip;}} (H, 2)-可求和为 <sup>1</sup>⁄<sub>4</sub>。
切萨罗求和有两种有名的广义化:让这些在概念上更简单的是(H, ''n'')法的序列,其中''n''为[[自然数]]。(H, 1)和为切萨罗求和,更高的方法则重复平均值的计算。在上文中,偶数平均值趋近于<sup>1</sup>⁄<sub>2</sub>,奇数平均值则全部等于0,所以平均值''的''平均值趋近于 0 与<sup>1</sup>⁄<sub>2</sub>的平均数,即<sup>1</sup>⁄<sub>4</sub>。<ref>Hardy, p.9. For the full details of the calculation, see Weidlich, pp.17–18.</ref>因此,{{nowrap|1 − 2 + 3 − 4 + &hellip;}}是(H, 2)-可求和为<sup>1</sup>⁄<sub>4</sub>。


符号“H”代表[[奥图·赫尔德]]。[[1882年]],他第一次证明了被现在数学家们所看作的在[[阿贝耳求和]]与 (H, ''n'') 求和之间的关系;{{nowrap|1 − 2 + 3 − 4 + &hellip;}} 是第一个例子。<ref>Ferraro, p.118; Tucciarone, p.10. Ferraro crit{{nowrap|1 − 2 + 3 − 4 + &hellip;}} icizes Tucciarone's explanation (p.7) of how Hölder himself thought of the general result, but the two authors' explanations of Hölder's treatment of 1 − 2 + 3 − 4 + &hellip; are similar.</ref><sup>1</sup>⁄<sub>4</sub>是 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} (H, 2) 和这个事实也保证了它是阿贝耳和;这些都将在下文直接予以证明。
符号“H”代表[[奥图·赫尔德]]。[[1882年]],他第一次证明了被现在数学家们所看作的在[[阿贝耳求和]]与(H, ''n'')求和之间的关系;{{nowrap|1 − 2 + 3 − 4 + &hellip;}}是第一个例子。<ref>Ferraro, p.118; Tucciarone, p.10. Ferraro crit{{nowrap|1 − 2 + 3 − 4 + &hellip;}}icizes Tucciarone's explanation (p.7) of how Hölder himself thought of the general result, but the two authors' explanations of Hölder's treatment of 1 − 2 + 3 − 4 + &hellip; are similar.</ref><sup>1</sup>⁄<sub>4</sub>是{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的(H, 2)和这个事实也保证了它是阿贝耳和;这些都将在下文直接予以证明。


另外一个普遍明确的切萨罗求和的广义化,是 (C, ''n'') 法的序列。已经证明了 (C, ''n'') 求和与 (H, ''n'') 求和均能给出相同的结果,但是它们却有不同的历史背景。在[[1887年]],切萨罗已经接近于陈述出 (C, ''n'') 求和的定义了,但是他只给出了少量的例子。特别的,他在计算 {{nowrap|1 − 2 + 3 − 4 + &hellip;,}} <sup>1</sup>⁄<sub>4</sub> 时所采用的方法可能是 (C, ''n'') 的另一种描述,但是在当时并没有对其进行证明。他在1890年正式定义了 (C, n) 法,以陈述他的定理:一个 (C, ''n'')-可求和级数与一个 (C, ''m'')-可求和级数的柯西乘积是 (C, ''m'' + ''n'' + 1)-可求和。<ref>Ferraro, pp.123–128.</ref>
另外一个普遍明确的切萨罗求和的广义化,是(C, ''n'')法的序列。已经证明了(C, ''n'')求和与(H, ''n'')求和均能给出相同的结果,但是它们却有不同的历史背景。在[[1887年]],切萨罗已经接近于陈述出(C, ''n'')求和的定义了,但是他只给出了少量的例子。特别的,他在计算{{nowrap|1 − 2 + 3 − 4 + &hellip;,}}为<sup>1</sup>⁄<sub>4</sub>时所采用的方法可能是(C, ''n'')的另一种描述,但是在当时并没有对其进行证明。他在1890年正式定义了(C, n)法,以陈述他的定理:一个(C, ''n'')-可求和级数与一个(C, ''m'')-可求和级数的柯西乘积是(C, ''m'' + ''n'' + 1)-可求和。<ref>Ferraro, pp.123–128.</ref>


=== 阿贝耳求和 ===
=== 阿贝耳求和 ===
第145行: 第145行:
在一份[[1749年]]的报告中,[[莱昂哈德·欧拉]]承认了级数发散,但准备用任何方式对其求和:
在一份[[1749年]]的报告中,[[莱昂哈德·欧拉]]承认了级数发散,但准备用任何方式对其求和:


{{Cquote 2|1=<small>……当该级数 1−2+3−4+5−6 等 的和为 <sup>1</sup>⁄<sub>4</sub> 时,那肯定出现了悖论。对该级数的100项相加,我们得到了–50,但是,101项的和却给出+51,这与 <sup>1</sup>⁄<sub>4</sub> 是截然不同的,而且这种差距还会随着项数增加而变得更大。不过我在前一段时间已经注意到了,有必要给“和”这个词赋予一个更加广泛的意义……。</small><ref>Euler et al, p.2. Although the paper was written in 1749, it was not published until 1768.</ref>}}
{{Cquote 2|1=<small>……当该级数 1−2+3−4+5−6 等 的和为<sup>1</sup>⁄<sub>4</sub>时,那肯定出现了悖论。对该级数的100项相加,我们得到了–50,但是,101项的和却给出+51,这与<sup>1</sup>⁄<sub>4</sub>是截然不同的,而且这种差距还会随着项数增加而变得更大。不过我在前一段时间已经注意到了,有必要给“和”这个词赋予一个更加广泛的意义……。</small><ref>Euler et al, p.2. 虽然这张纸写于1749年,但直到1768年才发表。</ref>}}


欧拉曾几次提议将“和”这个词广义化;参见[[欧拉与无穷级数]](''[[:en:Euler on infinite series|Euler on infinite series]]'')。在 {{nowrap|1 − 2 + 3 − 4 + &hellip;,}} 的情况下,他的设想与现在所知的[[阿贝耳求和]]相似:
欧拉曾几次提议将“和”这个词广义化;参见[[欧拉与无穷级数]](''[[:en:Euler on infinite series|Euler on infinite series]]'')。在{{nowrap|1 − 2 + 3 − 4 + &hellip;,}}的情况下,他的设想与现在所知的[[阿贝耳求和]]相似:


{{Cquote 2|1=<small>……毫无疑问,级数 1−2+3−4+5+更多 的和为 <sup>1</sup>⁄<sub>4</sub>;由于它是由公式 <sup>1</sup>⁄<sub>(1+1)<sup>2</sup></sub> 展开而成,而此公式的值明显为 <sup>1</sup>⁄<sub>4</sub>。在考虑一般级数 1 − 2''x'' + 3''x''<sup>2</sup> − 4''x''<sup>3</sup> + 5''x''<sup>4</sup> − 6''x''<sup>5</sup> + ''&c''后这个概念变得更明晰了。这个一般级数是由表达式 <sup>1</sup>⁄<sub>(1+''x'')<sup>2</sup></sub> 展开而成,当我们让 ''x'' = 1 后,这个级数就确确实实地相等了。</small><ref>Euler et al, pp.3, 25.</ref>}}
{{Cquote 2|1=<small>……毫无疑问,级数 1−2+3−4+5+更多 的和为<sup>1</sup>⁄<sub>4</sub>;由于它是由公式<sup>1</sup>⁄<sub>(1+1)<sup>2</sup></sub>展开而成,而此公式的值明显为<sup>1</sup>⁄<sub>4</sub>。在考虑一般级数 1 − 2''x'' + 3''x''<sup>2</sup> − 4''x''<sup>3</sup> + 5''x''<sup>4</sup> − 6''x''<sup>5</sup> + ''&c''后这个概念变得更明晰了。这个一般级数是由表达式<sup>1</sup>⁄<sub>(1+''x'')<sup>2</sup></sub>展开而成,当我们让 ''x'' = 1 后,这个级数就确确实实地相等了。</small><ref>Euler et al, pp.3, 25.</ref>}}


在当[[绝对值]] |''x''| < 1 时,有许多方式去验证欧拉的下列等式正确:
在当[[绝对值]] |''x''| < 1 时,有许多方式去验证欧拉的下列等式正确:
:<math>1-2x+3x^2-4x^3+\cdots = \frac{1}{(1+x)^2}.</math>
:<math>1-2x+3x^2-4x^3+\cdots = \frac{1}{(1+x)^2}.</math>


可以将[[泰勒级数|泰勒展开式]]的右边拿掉,或使用正规的[[多项式长除]]([[:en:polynomial long division|long division process for polynomials]])。从左方开始,可采用上文的一般启发式并尝试乘以两次 (1+''x''),或对几何级数 {{nowrap|1 − ''x'' + ''x''<sup>2</sup> − &hellip;}} 求平方。欧拉似乎也提出可以对后者级数的每项求[[微分]]。<ref>例如,Lavine (p.23)提倡长除但并没有得出结果;Vretblad (p.231)计算了柯西乘积。欧拉的建议是含糊的;参看Euler et al, pp.3, 26。 [[约翰·卡洛斯·贝兹|约翰·贝兹]]([[:en:John C. Baez|John Baez]])甚至提出一种包括将[[点集]]与[[量子谐振子]]相乘的范畴理论法。Baez, John C. [http://math.ucr.edu/home/baez/qg-winter2004/zeta.pdf 欧拉对 1 + 2 + 3 + &hellip; = 1/12 的证明 (PDF)] math.ucr.edu ([[2003年]][[12月19日]])。 [[2007年]][[3月11日]]检索。</ref>
可以将[[泰勒级数|泰勒展开式]]的右边拿掉,或使用正规的[[多项式长除]]([[:en:polynomial long division|long division process for polynomials]])。从左方开始,可采用上文的一般启发式并尝试乘以两次(1+''x''),或对几何级数{{nowrap|1 − ''x'' + ''x''<sup>2</sup> − &hellip;}}求平方。欧拉似乎也提出可以对后者级数的每项求[[微分]]。<ref>例如,Lavine (p.23)提倡长除但并没有得出结果;Vretblad (p.231)计算了柯西乘积。欧拉的建议是含糊的;参看Euler et al, pp.3, 26。 [[约翰·卡洛斯·贝兹|约翰·贝兹]]([[:en:John C. Baez|John Baez]])甚至提出一种包括将[[点集]]与[[量子谐振子]]相乘的范畴理论法。Baez, John C. [http://math.ucr.edu/home/baez/qg-winter2004/zeta.pdf 欧拉对 1 + 2 + 3 + &hellip; = 1/12 的证明 (PDF)] math.ucr.edu ([[2003年]][[12月19日]])。 [[2007年]][[3月11日]]检索。</ref>


以现代的眼光看,级数 1 − 2''x'' + 3''x''<sup>2</sup> − 4''x''<sup>3</sup> + &hellip; 并没有定义一个在 {{nowrap|1=''x'' = 1}} 时的[[函数]],因此其值不能简单地被替换为结果表达式。由于函数被定义为满足所有的 {{nowrap|&#124;''x''&#124; < 1}},所以仍可取得''x''趋近于1的极限,而这就是阿贝耳和的定义:
以现代的眼光看,级数 1 − 2''x'' + 3''x''<sup>2</sup> − 4''x''<sup>3</sup> + &hellip; 并没有定义一个在{{nowrap|1=''x'' = 1}}时的[[函数]],因此其值不能简单地被替换为结果表达式。由于函数被定义为满足所有的{{nowrap|&#124;''x''&#124; < 1}},所以仍可取得''x''趋近于1的极限,而这就是阿贝耳和的定义:


:<math>\lim_{x\rightarrow 1^{-}}\sum_{n=1}^\infty n(-x)^{n-1} = \lim_{x\rightarrow 1^{-}}\frac{1}{(1+x)^2} = \frac14.</math>
:<math>\lim_{x\rightarrow 1^{-}}\sum_{n=1}^\infty n(-x)^{n-1} = \lim_{x\rightarrow 1^{-}}\frac{1}{(1+x)^2} = \frac14.</math>


=== 欧拉与波莱尔 ===
=== 欧拉与波莱尔 ===
[[Image:Pm1234-Euler.png|right|thumb|<sup>1</sup>⁄<sub>2</sub> <sup>1</sup>⁄<sub>4</sub> 的欧拉求和]]
[[Image:Pm1234-Euler.png|right|thumb|<sup>1</sup>⁄<sub>2</sub>−<sup>1</sup>⁄<sub>4</sub>的欧拉求和]]
欧拉对该级数还使用了另外一种技巧:[[二项式变换#欧拉变换|欧拉变换]],这是他自己的发明。要计算欧拉变换,首先要有可形成交错级数的正项序列——在此情况下为 {{nowrap|1, 2, 3, 4, &hellip;}}。将此序列中的首项标示为 ''a''<sub>0</sub>。
欧拉对该级数还使用了另外一种技巧:[[二项式变换#欧拉变换|欧拉变换]],这是他自己的发明。要计算欧拉变换,首先要有可形成交错级数的正项序列——在此情况下为{{nowrap|1, 2, 3, 4, &hellip;}}。将此序列中的首项标示为 ''a''<sub>0</sub>。


下一步需要 {{nowrap|1, 2, 3, 4, &hellip;}} 的[[差分|前向差分]];这恰好是 {{nowrap|1, 1, 1, 1, &hellip;}}。将该序列的首项标示为 Δ''a''<sub>0</sub>。欧拉变换也基于差分的差分,以及更高的叠函数,但是 {{nowrap|1, 1, 1, 1, &hellip;}} 的前向差分为0。{{nowrap|1 − 2 + 3 − 4 + &hellip;}} 的欧拉变换便可定义为:
下一步需要{{nowrap|1, 2, 3, 4, &hellip;}}的[[差分|前向差分]];这恰好是{{nowrap|1, 1, 1, 1, &hellip;}}。将该序列的首项标示为 Δ''a''<sub>0</sub>。欧拉变换也基于差分的差分,以及更高的叠函数,但是{{nowrap|1, 1, 1, 1, &hellip;}}的前向差分为0。{{nowrap|1 − 2 + 3 − 4 + &hellip;}}的欧拉变换便可定义为:
:<math>\frac12 a_0-\frac14\Delta a_0 +\frac18\Delta^2 a_0 -\cdots = \frac12-\frac14.</math>
:<math>\frac12 a_0-\frac14\Delta a_0 +\frac18\Delta^2 a_0 -\cdots = \frac12-\frac14.</math>
用现代术语来说,{{nowrap|1 − 2 + 3 − 4 + &hellip;}} 是[[欧拉求和|欧拉可求和]]并为 <sup>1</sup>⁄<sub>4</sub>。
用现代术语来说,{{nowrap|1 − 2 + 3 − 4 + &hellip;}}是[[欧拉求和|欧拉可求和]]并为<sup>1</sup>⁄<sub>4</sub>。


欧拉可求和也包含有另一种可求和法。将 {{nowrap|1 − 2 + 3 − 4 + &hellip;}} 表示为:
欧拉可求和也包含有另一种可求和法。将{{nowrap|1 − 2 + 3 − 4 + &hellip;}}表示为:
:<math>\sum_{k=0}^\infty a_k = \sum_{k=0}^\infty(-1)^k(k+1),</math>
:<math>\sum_{k=0}^\infty a_k = \sum_{k=0}^\infty(-1)^k(k+1),</math>


第178行: 第178行:


=== 比例分离 ===
=== 比例分离 ===
赛切夫与Woyczyński只通过两个物理原理便得出了 {{nowrap|1=1 − 2 + 3 − 4 + &hellip; = <sup>1</sup>⁄<sub>4</sub>}},这两个原理分别是:近零张弛(''infinitesimal relaxation'')与比例分离(''separation of scales'')。为了表示得精确,他们为这些原理定义了一系列的“''φ''-求和法”,所有这些方法都可以将级数求和得 <sup>1</sup>⁄<sub>4</sub>:
赛切夫与Woyczyński只通过两个物理原理便得出了{{nowrap|1=1 − 2 + 3 − 4 + &hellip; =<sup>1</sup>⁄<sub>4</sub>}},这两个原理分别是:近零张弛(''infinitesimal relaxation'')与比例分离(''separation of scales'')。为了表示得精确,他们为这些原理定义了一系列的“''φ''-求和法”,所有这些方法都可以将级数求和得<sup>1</sup>⁄<sub>4</sub>:


*如果''φ''(''x'')是一个函数,其一、二阶导数在(0, ∞)上是连续可积分的,这样的话φ(0) = 1 ,并且φ(''x'')的极限与''x''φ(''x'')在+∞时的值均为0,然后:<ref>Saichev and Woyczyński, pp.260–264.</ref>
*如果''φ''(''x'')是一个函数,其一、二阶导数在(0, ∞)上是连续可积分的,这样的话φ(0) = 1 ,并且φ(''x'')的极限与''x''φ(''x'')在+∞时的值均为0,然后:<ref>Saichev and Woyczyński, pp.260–264.</ref>
第188行: 第188行:
==广义化==
==广义化==
[[Image:Pm1234 Euler1755 I-V.png|thumb|1755年的《''Institutiones''》上,欧拉对相似的级数求和]]
[[Image:Pm1234 Euler1755 I-V.png|thumb|1755年的《''Institutiones''》上,欧拉对相似的级数求和]]
{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的三倍柯西积为{{nowrap|1 − 3 + 6 − 10 + &hellip;}},为[[三角形数]]([[:en:triangular number|triangular number]]s)的交错级数;其阿贝耳与欧拉和为<sup>1</sup>⁄<sub>8</sub>。<ref>Kline, p.313.</ref>{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的四倍柯西积为{{nowrap|1 − 4 + 10 − 20 + &hellip;}},为[[四面体数]]([[:en:tetrahedral number|tetrahedral number]]s)的交错级数,这个的阿贝耳和为<sup>1</sup>⁄<sub>16</sub>。
{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的三倍柯西积为{{nowrap|1 − 3 + 6 − 10 + &hellip;}},为[[三角形数]]的交错级数;其阿贝耳与欧拉和为<sup>1</sup>⁄<sub>8</sub>。<ref>Kline, p.313.</ref>{{nowrap|1 − 1 + 1 − 1 + &hellip;}}的四倍柯西积为{{nowrap|1 − 4 + 10 − 20 + &hellip;}},为[[四面体数]]的交错级数,这个的阿贝耳和为<sup>1</sup>⁄<sub>16</sub>。


另一个1 − 2 + 3 − 4 + &hellip;在略微不同的方向的广义化是级数{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}},使用了另外的值''n''。对正整数''n''来说,此级数有下列的阿贝耳和:<ref>Knopp, p.491; there appears to be an error at this point in Hardy, p.3.</ref>
另一个1 − 2 + 3 − 4 + &hellip;在略微不同的方向的广义化是级数{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}},使用了另外的值''n''。对正整数''n''来说,此级数有下列的阿贝耳和:<ref>Knopp, p.491; there appears to be an error at this point in Hardy, p.3.</ref>
第201行: 第201行:
:其中,''n''为正数。这是一个笑料,朋友。”<ref>Grattan-Guinness, p.80. See Markushevich, p.48, for a different translation from the original [[French language|French]]; the tone remains the same.</ref>
:其中,''n''为正数。这是一个笑料,朋友。”<ref>Grattan-Guinness, p.80. See Markushevich, p.48, for a different translation from the original [[French language|French]]; the tone remains the same.</ref>


切萨罗的老师[[欧仁·查理·卡塔兰]]([[:en:Eugène Charles Catalan|Eugène Charles Catalan]])也轻视发散级数。在卡塔兰的影响下,切萨罗早期提出{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}}的“习用式”是“荒谬的等式”;而在[[1883年]],切萨罗表明了当时的一个典型看法:公式是错的,不过在某些场合在形式上是有用的。最后,在他[[1890年]]的书《''Sur la multiplication des séries''》中,切萨罗使用了一个接近于定义的模型。<ref>Ferraro, pp.120–128.</ref>
切萨罗的老师[[欧仁·查理·卡塔兰]]也轻视发散级数。在卡塔兰的影响下,切萨罗早期提出{{nowrap|1 − 2<sup>''n''</sup> + 3<sup>''n''</sup> − 4<sup>''n''</sup> + &hellip;}}的“习用式”是“荒谬的等式”;而在[[1883年]],切萨罗表明了当时的一个典型看法:公式是错的,不过在某些场合在形式上是有用的。最后,在他[[1890年]]的书《''Sur la multiplication des séries''》中,切萨罗使用了一个接近于定义的模型。<ref>Ferraro, pp.120–128.</ref>


此级数亦研究过''n''为非线性值的情况;这产生了狄利克雷η函数。欧拉研究{{nowrap|1 − 2 + 3 − 4 + &hellip;}}相关级数的部分动机是η函数的[[泛函方程]],这直接导向了[[黎曼ζ函数]]的泛函方程。欧拉在正[[奇数和偶数|偶整数]](包括在[[巴塞尔问题]]中)时找到这些函数值的建树已让他闻名世界,他也试图找到正[[奇数和偶数|奇整数]](包括在[[:en:Apéry's constant|Apéry's constant]]中) 时的值,但这个问题直到今天都是令人困惑的。η函数通过欧拉的方法解决会更加简单,因为它的[[狄利克雷级数]]是处处阿贝耳可求和;η函数的狄利克雷级数非常难以对发散的部分求和。<ref>Euler et al, pp.20–25.</ref>例如,{{nowrap|1 − 2 + 3 − 4 + &hellip;}}在η函数中的相似级数是非交错级数{{nowrap|[[1 + 2 + 3 + 4 + · · ·|1 + 2 + 3 + 4 + &hellip;]]}},该级数在现代[[物理学]]上有很深的应用,不过需要非常强的方法才能求和。
此级数亦研究过''n''为非线性值的情况;这产生了狄利克雷η函数。欧拉研究{{nowrap|1 − 2 + 3 − 4 + &hellip;}}相关级数的部分动机是η函数的[[泛函方程]],这直接导向了[[黎曼ζ函数]]的泛函方程。欧拉在正[[奇数和偶数|偶整数]](包括在[[巴塞尔问题]]中)时找到这些函数值的建树已让他闻名世界,他也试图找到正[[奇数和偶数|奇整数]](包括在[[:en:Apéry's constant|Apéry's constant]]中) 时的值,但这个问题直到今天都是令人困惑的。η函数通过欧拉的方法解决会更加简单,因为它的[[狄利克雷级数]]是处处阿贝耳可求和;η函数的狄利克雷级数非常难以对发散的部分求和。<ref>Euler et al, pp.20–25.</ref>例如,{{nowrap|1 − 2 + 3 − 4 + &hellip;}}在η函数中的相似级数是非交错级数{{nowrap|[[1 + 2 + 3 + 4 + · · ·|1 + 2 + 3 + 4 + &hellip;]]}},该级数在现代[[物理学]]上有很深的应用,不过需要非常强的方法才能求和。


== 注解 ==
{{reflist|2}}


==参考书目==
<div class="references-small">
*{{cite book |last=Beals |first=Richard |title=Analysis: an introduction |year=2004 |publisher=Cambridge UP |isbn= 0-521-60047-2}}
*{{cite book |last=Davis |first=Harry F. |title=Fourier Series and Orthogonal Functions |year=1989 |month=May |publisher=Dover |isbn= 0-486-65973-9}}
*{{cite web |author=Euler, Leonhard; Lucas Willis; and Thomas J Osler |title=Translation with notes of Euler's paper: Remarks on a beautiful relation between direct as well as reciprocal power series |year=2006 |publisher=The Euler Archive |url=http://www.math.dartmouth.edu/~euler/pages/E352.html |accessdate=2007-03-22}} Originally published as {{cite journal |last=Euler |first=Leonhard |title=Remarques sur un beau rapport entre les séries des puissances tant directes que réciproques |journal=Memoires de l'academie des sciences de Berlin |year=1768 |volume=17 |pages=83–106}}
*{{cite journal |last=Ferraro |first=Giovanni |title=The First Modern Definition of the Sum of a Divergent Series: An Aspect of the Rise of 20th Century Mathematics |journal=Archive for History of Exact Sciences |year=1999 |month=June |volume=54 |issue=2 |pages=101–135 |id={{doi|10.1007/s004070050036}}}}
*{{cite book |last=Grattan-Guinness |authorlink=Ivor Grattan-Guinness |first=Ivor |year=1970 |title=The development of the foundations of mathematical analysis from Euler to Riemann |publisher=MIT Press |isbn= 0-262-07034-0}}
*{{cite book |last=Hardy |first=G.H. |authorlink=G. H. Hardy |title=Divergent Series |year=1949 |publisher=Clarendon Press |id={{LCCN|91|0|75377}}}}
*{{cite journal |last=Kline |first=Morris |authorlink=Morris Kline |title=Euler and Infinite Series |journal=Mathematics Magazine |volume=56 |issue=5 |year=1983 |month=November |pages=307–314 |url=http://links.jstor.org/sici?sici=0025-570X%28198311%2956%3A5%3C307%3AEAIS%3E2.0.CO%3B2-M}}
*{{cite book |first=Shaughan |last=Lavine |title=Understanding the Infinite |year=1994 |publisher=Harvard UP |isbn= 0674920961}}
*{{cite book |last=Markushevich |first=A.I. |title=Series: fundamental concepts with historical exposition |year=1967 |edition=English translation of 3rd revised edition (1961) in Russian |publisher=Hindustan Pub. Corp. |id={{LCCN|68|0|17528}}}}
*{{cite book |author=Saichev, A.I., and W.A. Woyczyński |title=Distributions in the physical and engineering sciences, Volume 1 |publisher=Birkhaüser |year=1996 |isbn= 0-8176-3924-1}}
*{{cite journal |last=Tucciarone |first=John |title=The development of the theory of summable divergent series from 1880 to 1925 |journal=Archive for History of Exact Sciences |volume=10 |issue=1-2 |year=1973 |month=January |pages=1–40 |id={{doi|10.1007/BF00343405}}}}
*{{cite book |first=Anders |last=Vretblad |title=Fourier Analysis and Its Applications |year=2003 |publisher=Springer |isbn= 0387008365}}
*{{cite book |last=Weidlich |first=John E. |title=Summability methods for divergent series |year=1950 |month=June |publisher=Stanford M.S. theses|id={{OCLC|38624384}}}}
</div>

[[Category:数学级数]]
[[en:1 − 2 + 3 − 4 + · · ·]]
[[en:1 − 2 + 3 − 4 + · · ·]]
[[es:1 − 2 + 3 − 4 + · · ·]]
[[es:1 − 2 + 3 − 4 + · · ·]]

2007年4月20日 (五) 04:51的版本

1 − 2 + 3 − 4 + …的前几千项相加结果示意图

数学中,1 − 2 + 3 − 4 + …为首项为1,每项正负号交替出现且为连续正整数无穷级数。使用Σ符号表示前m项之和可写作:

此无穷级数发散,即其部分和的序列(1, −1, 2, −2, …)不会趋近于任一有穷极限。此可等价地认为1 − 2 + 3 − 4 + …不存在和。

不过,在18世纪中期,莱昂哈德·欧拉写出了一个他承认为悖论等式

该等式的严密解释在很久以后才出现。1890年初,恩纳斯托·切萨罗埃米尔·博雷尔与其他一些数学家研究出了定义明确well-defined)的方法,来为有穷极限确定广义级数的和——其中包含了欧拉所寻找的新解释。大部分这些可求和法最终可简单地确定1 − 2 + 3 − 4 + …的“和”为14切萨罗求和Cesàro summation)是少数几种不能计算出1 − 2 + 3 − 4 + …之和的方法。因此此级数是需要某个略强的方法——譬如阿贝耳求和Abel summation)——的例子。

级数1, −1, 2, −2, …格兰第级数Grandi's series1 − 1 + 1 − 1 + …有较近的联系。欧拉将此二级数视作1 − 2n + 3n − 4n + …对任意n的特殊情况;这是他在巴塞尔问题Basel problem)上所作工作的一个延伸研究方向,并导向了泛函方程——即我们现在所知的狄利克雷η函数黎曼ζ函数

发散

级数项(1, −2, 3, −4, …)没有趋近于0;因此可通过项测试term test)来确定1 − 2 + 3 − 4 + …发散。作为后文的参考,此方法也常被用于从基础级上预见发散。从定义可知,无穷级数的收敛或发散是由其部分和的收敛或发散确定的,1 − 2 + 3 − 4 + …的部分和为:[1]

1 = 1,
1 − 2 = −1,
1 − 2 + 3 = 2,
1 − 2 + 3 − 4 = −2,
1 − 2 + 3 − 4 + 5 = 3,
1 − 2 + 3 − 4 + 5 − 6 = −3,

此序列的一个显著的特点是每个整数都出现了一次——如果将空部分和计入还包括0——因此其可数集为整数集[2]很明显的,不可能让变化的结果辐合到一个确定的数,因此1 − 2 + 3 − 4 + …发散。

求和的启发

稳定性与线性

由于各项 1, −2, 3, −4, 5, −6, … 以一种简单模式排列,级数1 − 2 + 3 − 4 + …可表示为它自己的变换形式(概略的说法),而因此所得的方程解取得一数值。暂时假设写作s = 1 − 2 + 3 − 4 + …有意义——其中的s为常数,然后再处理s =14:的问题:

复制4份 1 − 2 + 3 − 4 + …,仅使用移动与项项相加,结果为1。左右两边的两个1 − 2 + 3 − 4 + …副本相互抵消,并得出1 − 1 + 1 − 1 + …
4 s = (1 − 2 + 3 − 4 + …) + (1 − 2 + 3 − 4 + …) + (1 − 2 + 3 − 4 + …) + (1 − 2 + 3 − 4 + …)
= (1 − 2 + 3 − 4 + …) + 1 + (−2 + 3 − 4 + 5 - …) + 1 + (−2 + 3 − 4 + 5 - …) − 1 + (3 − 4 + 5 - 6 + …)
= 1 + ((1 - 2 - 2 + 3) + (-2 + 3 + 3 - 4) + (3 - 4 - 4 + 5) + (-4 + 5 + 5 - 6) + …)
4 s = 1 + (0 + 0 + 0 + 0 …) = 1,

因此,s =14.[3],如右图所示。

尽管1 − 2 + 3 − 4 + …没有通常意义的和,等式s = 1 − 2 + 3 − 4 + … =14却可被赋予另外一种意义。离散级数之“和”的一种普遍定义被称为一种求和法(summation method)或可和法(summability method)——对所有可能级数的一些子集求和。有许多种不同的方法(部分将在下文中出现),这些方法有着与常数求和所共有的特性。以上的处理实际上都可由下述所证明:给出任意的线形且稳定的可和法,并计算级数1 − 2 + 3 − 4 + …,结果为14。此外,由于:

2 s = (1 − 2 + 3 − 4 + …) + (1 − 2 + 3 − 4 + …)
= 1 + (−2 + 3 − 4 + …) + 1 - 2 + (3 − 4 + 5 - …)
= 0 + ((-2 + 3)   +   (3 - 4)   +   (-4 + 5)   +   (5 - 6)   +   …)
½ = 1 - 1 + 1 - 1 …,

故此方法也一定能对格兰第级数求和,并得结果为1 − 1 + 1 − 1 + … =12

以上为1 − 2 + 3 − 4 + …的一般和限制了可能的值,但依然不能首先显现出哪种方法能或不能计算出此级数来。事实上,一些线形且稳定的可和法,譬如常数求和,是无法计算出1 − 2 + 3 − 4 + …的。看待此级数的不同眼光有助于确定哪些方法可计算出14:表现为的形式。

柯西乘积

1891年,, 恩纳斯托·切萨罗Ernesto Cesàro)表示了将分散级数严密地带入微积分学的愿望,并指出:“已可写出(1 − 1 + 1 − 1 + …)2 = 1 − 2 + 3 − 4 + …并断定两边均等于14。”[4]对切萨罗而言,这个等式是他前几年发表的一个定理的应用,该定理也许是历史上可求和的发散级数的第一个定理。关于此求和法的详细内容请见下文;其中心思想是1 − 2 + 3 − 4 + …1 − 1 + 1 − 1 + …1 − 1 + 1 − 1 + …柯西乘积Cauchy product)。

File:Pm1234 Cauchy.svg
1 − 2 + 3 − 4 + … 以 1 − 1 + 1 − 1 + … 的二重柯西乘积出现

两个无穷级数的柯西乘积可被确定,即使在他们都发散的时候。在 Σan= Σbn= Σ(−1)n 的情况下,柯西乘积的项可由有穷对角线求和的方式给出:

积级数为:

这样一种考虑到两个级数的柯西乘积与1 − 1 + 1 − 1 + … =12结果的求和法,也能够求出1 − 2 + 3 − 4 + … =14。由前部分的结果可知,当方法是线形、稳定并考虑到柯西乘积的时候,1 − 1 + 1 − 1 + …1 − 2 + 3 − 4 + …的可求和之间是等价的。

切萨罗的定理是一个深奥的例子。级数1 − 1 + 1 − 1 + …在最弱的意义上是切萨罗可求和,称作(C, 1)-可求和,然而1 − 2 + 3 − 4 + …则需要切萨罗的定理的一个更强的形式[5],表示为(C, 2)-可求和。由于切萨罗的定理的所有形式均为线形且稳定的,所得的值正是此前计算所得的。

特殊方法

切萨罗与赫尔德

关于14的(H, 2)和的数据

要找到1 − 2 + 3 − 4 + &hellip的(C, 1)切萨罗和,如果它存在,就需要计算该级数部分和的算术平均值。 部分和为:

1, −1, 2, −2, 3, −3, …,

这些部分和的算术平均值为:

1, 0,23, 0,35, 0,47, ….

此平均值序列没有辐合,因此 1 − 2 + 3 − 4 + … 不是切萨罗可求和。

切萨罗求和有两种有名的广义化:让这些在概念上更简单的是(H, n)法的序列,其中n自然数。(H, 1)和为切萨罗求和,更高的方法则重复平均值的计算。在上文中,偶数平均值趋近于12,奇数平均值则全部等于0,所以平均值平均值趋近于 0 与12的平均数,即14[6]因此,1 − 2 + 3 − 4 + …是(H, 2)-可求和为14

符号“H”代表奥图·赫尔德1882年,他第一次证明了被现在数学家们所看作的在阿贝耳求和与(H, n)求和之间的关系;1 − 2 + 3 − 4 + …是第一个例子。[7]141 − 2 + 3 − 4 + …的(H, 2)和这个事实也保证了它是阿贝耳和;这些都将在下文直接予以证明。

另外一个普遍明确的切萨罗求和的广义化,是(C, n)法的序列。已经证明了(C, n)求和与(H, n)求和均能给出相同的结果,但是它们却有不同的历史背景。在1887年,切萨罗已经接近于陈述出(C, n)求和的定义了,但是他只给出了少量的例子。特别的,他在计算1 − 2 + 3 − 4 + …,14时所采用的方法可能是(C, n)的另一种描述,但是在当时并没有对其进行证明。他在1890年正式定义了(C, n)法,以陈述他的定理:一个(C, n)-可求和级数与一个(C, m)-可求和级数的柯西乘积是(C, m + n + 1)-可求和。[8]

阿贝耳求和

1−2x+3x2+…; 1/(1 + x)2 的一部分;其极限为1

在一份1749年的报告中,莱昂哈德·欧拉承认了级数发散,但准备用任何方式对其求和:

……当该级数 1−2+3−4+5−6 等 的和为14时,那肯定出现了悖论。对该级数的100项相加,我们得到了–50,但是,101项的和却给出+51,这与14是截然不同的,而且这种差距还会随着项数增加而变得更大。不过我在前一段时间已经注意到了,有必要给“和”这个词赋予一个更加广泛的意义……。[9]

欧拉曾几次提议将“和”这个词广义化;参见欧拉与无穷级数Euler on infinite series)。在1 − 2 + 3 − 4 + …,的情况下,他的设想与现在所知的阿贝耳求和相似:

……毫无疑问,级数 1−2+3−4+5+更多 的和为14;由于它是由公式1(1+1)2展开而成,而此公式的值明显为14。在考虑一般级数 1 − 2x + 3x2 − 4x3 + 5x4 − 6x5 + &c后这个概念变得更明晰了。这个一般级数是由表达式1(1+x)2展开而成,当我们让 x = 1 后,这个级数就确确实实地相等了。[10]

在当绝对值 |x| < 1 时,有许多方式去验证欧拉的下列等式正确:

可以将泰勒展开式的右边拿掉,或使用正规的多项式长除long division process for polynomials)。从左方开始,可采用上文的一般启发式并尝试乘以两次(1+x),或对几何级数1 − x + x2 − …求平方。欧拉似乎也提出可以对后者级数的每项求微分[11]

以现代的眼光看,级数 1 − 2x + 3x2 − 4x3 + … 并没有定义一个在x = 1时的函数,因此其值不能简单地被替换为结果表达式。由于函数被定义为满足所有的|x| < 1,所以仍可取得x趋近于1的极限,而这就是阿贝耳和的定义:

欧拉与波莱尔

1214的欧拉求和

欧拉对该级数还使用了另外一种技巧:欧拉变换,这是他自己的发明。要计算欧拉变换,首先要有可形成交错级数的正项序列——在此情况下为1, 2, 3, 4, …。将此序列中的首项标示为 a0

下一步需要1, 2, 3, 4, …前向差分;这恰好是1, 1, 1, 1, …。将该序列的首项标示为 Δa0。欧拉变换也基于差分的差分,以及更高的叠函数,但是1, 1, 1, 1, …的前向差分为0。1 − 2 + 3 − 4 + …的欧拉变换便可定义为:

用现代术语来说,1 − 2 + 3 − 4 + …欧拉可求和并为14

欧拉可求和也包含有另一种可求和法。将1 − 2 + 3 − 4 + …表示为:

就有了相关的处处收敛级数:

因此 1 − 2 + 3 − 4 + … 的波莱尔和为:[12]

比例分离

赛切夫与Woyczyński只通过两个物理原理便得出了1 − 2 + 3 − 4 + … =14,这两个原理分别是:近零张弛(infinitesimal relaxation)与比例分离(separation of scales)。为了表示得精确,他们为这些原理定义了一系列的“φ-求和法”,所有这些方法都可以将级数求和得14

  • 如果φ(x)是一个函数,其一、二阶导数在(0, ∞)上是连续可积分的,这样的话φ(0) = 1 ,并且φ(x)的极限与xφ(x)在+∞时的值均为0,然后:[13]

该结果推广了阿贝耳求和,当取φ(x) = exp(−x)时可得到先前的等式。此一般陈述可通过将关于m的级数中的项配对,并将表达式变换为黎曼积分的形式予以证明。在后一步中,对1 − 1 + 1 − 1 + …相应证明运用了中值定理,但在这里需要泰勒公式中更强的拉格朗日形式

广义化

1755年的《Institutiones》上,欧拉对相似的级数求和

1 − 1 + 1 − 1 + …的三倍柯西积为1 − 3 + 6 − 10 + …,为三角形数的交错级数;其阿贝耳与欧拉和为18[14]1 − 1 + 1 − 1 + …的四倍柯西积为1 − 4 + 10 − 20 + …,为四面体数的交错级数,这个的阿贝耳和为116

另一个1 − 2 + 3 − 4 + …在略微不同的方向的广义化是级数1 − 2n + 3n − 4n + …,使用了另外的值n。对正整数n来说,此级数有下列的阿贝耳和:[15]

其中Bn伯努利数。对偶数n,则变为:

后一个和在1826年成为尼尔斯·亨利克·阿贝尔特别嘲笑的对象: 如果用到它们 想要的

“发散级数纯粹是魔鬼的工作,胆敢去找到任何证明它们的行为都是羞耻的。如果用到它们,可以从中获得想要的东西;同时也是它们,制造了如此多的不愉快与如此多的悖论。试问能想到比下面内容更令人惊恐的东西吗:
0 = 1 − 2n + 3n − 4n + etc.
其中,n为正数。这是一个笑料,朋友。”[16]

切萨罗的老师欧仁·查理·卡塔兰也轻视发散级数。在卡塔兰的影响下,切萨罗早期提出1 − 2n + 3n − 4n + …的“习用式”是“荒谬的等式”;而在1883年,切萨罗表明了当时的一个典型看法:公式是错的,不过在某些场合在形式上是有用的。最后,在他1890年的书《Sur la multiplication des séries》中,切萨罗使用了一个接近于定义的模型。[17]

此级数亦研究过n为非线性值的情况;这产生了狄利克雷η函数。欧拉研究1 − 2 + 3 − 4 + …相关级数的部分动机是η函数的泛函方程,这直接导向了黎曼ζ函数的泛函方程。欧拉在正偶整数(包括在巴塞尔问题中)时找到这些函数值的建树已让他闻名世界,他也试图找到正奇整数(包括在Apéry's constant中) 时的值,但这个问题直到今天都是令人困惑的。η函数通过欧拉的方法解决会更加简单,因为它的狄利克雷级数是处处阿贝耳可求和;η函数的狄利克雷级数非常难以对发散的部分求和。[18]例如,1 − 2 + 3 − 4 + …在η函数中的相似级数是非交错级数1 + 2 + 3 + 4 + …,该级数在现代物理学上有很深的应用,不过需要非常强的方法才能求和。

注解

  1. ^ Hardy p.8
  2. ^ Beals p.23
  3. ^ Hardy (p.6) 结合格兰第级数1 − 1 + 1 − 1 + …的计算提出了此推导过程。
  4. ^ "One already writes(1 − 1 + 1 − 1 + …)2 = 1 − 2 + 3 − 4 + …and asserts that both the sides are equal to14.", Ferraro, p.130.
  5. ^ Hardy, p.3; Weidlich, pp.52–55.
  6. ^ Hardy, p.9. For the full details of the calculation, see Weidlich, pp.17–18.
  7. ^ Ferraro, p.118; Tucciarone, p.10. Ferraro crit1 − 2 + 3 − 4 + …icizes Tucciarone's explanation (p.7) of how Hölder himself thought of the general result, but the two authors' explanations of Hölder's treatment of 1 − 2 + 3 − 4 + … are similar.
  8. ^ Ferraro, pp.123–128.
  9. ^ Euler et al, p.2. 虽然这张纸写于1749年,但直到1768年才发表。
  10. ^ Euler et al, pp.3, 25.
  11. ^ 例如,Lavine (p.23)提倡长除但并没有得出结果;Vretblad (p.231)计算了柯西乘积。欧拉的建议是含糊的;参看Euler et al, pp.3, 26。 约翰·贝兹John Baez)甚至提出一种包括将点集量子谐振子相乘的范畴理论法。Baez, John C. 欧拉对 1 + 2 + 3 + … = 1/12 的证明 (PDF)。 math.ucr.edu (2003年12月19日)。 2007年3月11日检索。
  12. ^ Weidlich p. 59
  13. ^ Saichev and Woyczyński, pp.260–264.
  14. ^ Kline, p.313.
  15. ^ Knopp, p.491; there appears to be an error at this point in Hardy, p.3.
  16. ^ Grattan-Guinness, p.80. See Markushevich, p.48, for a different translation from the original French; the tone remains the same.
  17. ^ Ferraro, pp.120–128.
  18. ^ Euler et al, pp.20–25.

参考书目

  • Beals, Richard. Analysis: an introduction. Cambridge UP. 2004. ISBN 0-521-60047-2. 
  • Davis, Harry F. Fourier Series and Orthogonal Functions. Dover. 1989. ISBN 0-486-65973-9.  已忽略未知参数|month=(建议使用|date=) (帮助)
  • Euler, Leonhard; Lucas Willis; and Thomas J Osler. Translation with notes of Euler's paper: Remarks on a beautiful relation between direct as well as reciprocal power series. The Euler Archive. 2006 [2007-03-22].  Originally published as Euler, Leonhard. Remarques sur un beau rapport entre les séries des puissances tant directes que réciproques. Memoires de l'academie des sciences de Berlin. 1768, 17: 83–106. 
  • Ferraro, Giovanni. The First Modern Definition of the Sum of a Divergent Series: An Aspect of the Rise of 20th Century Mathematics. Archive for History of Exact Sciences. 1999, 54 (2): 101–135. doi:10.1007/s004070050036.  已忽略未知参数|month=(建议使用|date=) (帮助)
  • Grattan-Guinness, Ivor. The development of the foundations of mathematical analysis from Euler to Riemann. MIT Press. 1970. ISBN 0-262-07034-0. 
  • Hardy, G.H. Divergent Series. Clarendon Press. 1949.
    LCCN 91-0 – 0
    .
     
  • Kline, Morris. Euler and Infinite Series. Mathematics Magazine. 1983, 56 (5): 307–314.  已忽略未知参数|month=(建议使用|date=) (帮助)
  • Lavine, Shaughan. Understanding the Infinite. Harvard UP. 1994. ISBN 0674920961. 
  • Markushevich, A.I. Series: fundamental concepts with historical exposition English translation of 3rd revised edition (1961) in Russian. Hindustan Pub. Corp. 1967.
    LCCN 68-0 – 0
    .
     
  • Saichev, A.I., and W.A. Woyczyński. Distributions in the physical and engineering sciences, Volume 1. Birkhaüser. 1996. ISBN 0-8176-3924-1. 
  • Tucciarone, John. The development of the theory of summable divergent series from 1880 to 1925. Archive for History of Exact Sciences. 1973, 10 (1-2): 1–40. doi:10.1007/BF00343405.  已忽略未知参数|month=(建议使用|date=) (帮助)
  • Vretblad, Anders. Fourier Analysis and Its Applications. Springer. 2003. ISBN 0387008365. 
  • Weidlich, John E. Summability methods for divergent series. Stanford M.S. theses. 1950. OCLC 38624384.  已忽略未知参数|month=(建议使用|date=) (帮助)