关于Euler-Poisson积分的几种解法

来源:https://www.cnblogs.com/Renascence-5/p/5432211.html

方法1:因为积分值只与被积函数和积分域有关,与积分变量无关,所以
[I^{2}=left ( int_{0}^{infty }e^{-x^{2}}mathrm{d}x ight )^{2}=int_{0}^{infty }e^{-x^{2}}mathrm{d}x~~cdot int_{0}^{infty }e^{-y^{2}}mathrm{d}y=int_{0}^{infty }int_{0}^{infty }e^{-left ( x^{2}+y^{2} ight )}mathrm{d}xmathrm{d}y]
用极坐标系下二重积分的计算法
[I=int_{0}^{frac{pi }{2}}mathrm{d} heta int_{0}^{infty }e^{-r^{2}}rmathrm{d}r=frac{pi }{2}left ( -frac{1}{2}e^{-r^{2}} ight )Bigg|_{0}^{infty }=frac{pi }{4}]
而 (e^{-r^{2}}geq 0), 则 (I>0). 即
[I=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=sqrt{frac{pi }{4}}=frac{sqrt{pi }}{2}]
方法2:因为 (displaystyle left ( 1+frac{x^{2}}{n} ight )^{-n}) 当 $n ightarrow +infty $ 时一致收敛于 (e^{-x^{2}}), 利用积分号下取极限,则有
[I=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=int_{0}^{infty }left [ lim_{n ightarrow infty }left ( 1+frac{x^{2}}{n} ight )^{-n} ight ]mathrm{d}x=lim_{n ightarrow infty }int_{0}^{infty }left ( 1+frac{x^{2}}{n} ight )^{-n}mathrm{d}x]
令 (x=sqrt{nt}), 则
[I=lim_{n ightarrow infty }sqrt{n}int_{0}^{infty }frac{1}{left ( 1+t^{2} ight )^{n}}mathrm{d}t=sqrt{n}I_n]
由于
[egin{align*} I_{n-1}&=int_{0}^{infty }frac{1}{left ( 1+t^{2} ight )^{n-1}}mathrm{d}t=frac{t}{left ( 1+t^{2} ight )^{n-1}}Bigg|_{0}^{infty }+2left ( n-1 ight )int_{0}^{infty }frac{1}{left ( 1+t^{2} ight )}mathrm{d}t\ &=2left ( n-1 ight )I_{n-1}-2left ( n-1 ight )I_n end{align*}]
所以 (displaystyle I_n=frac{2n-3}{2n-2}I_{n-1}), 而 (displaystyle I_1=int_{0}^{infty }frac{1}{1+t^{2}}mathrm{d}t=frac{pi }{2}), 递推得
[I_n=frac{left ( 2n-3 ight )!!}{left ( 2n-2 ight )!!}cdot frac{pi }{2}]
因此 (displaystyle int_{0}^{infty }e^{-x^{2}}mathrm{d}x=lim_{n ightarrow infty }frac{sqrt{n}left ( 2n-3 ight )!!}{left ( 2n-2 ight )!!}cdot frac{pi }{2}). 根据Wallis公式,有
[frac{pi }{2}=lim_{n ightarrow infty }frac{left [ left ( 2n ight )!! ight ]^{2}}{left ( 2n+1 ight )left [ left ( 2n-1 ight )!! ight ]^{2}}=lim_{n ightarrow infty }frac{left [ left ( 2n-2 ight )!! ight ]^{2}}{left ( 2n-1 ight )left [ left ( 2n-1 ight )!! ight ]^{2}}]
所以
[egin{align*} I&=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=frac{pi }{2}lim_{n ightarrow infty }frac{sqrt{n}left ( 2n-3 ight )!!}{left ( 2n-2 ight )!!}=frac{pi }{2}lim_{n ightarrow infty }frac{left ( 2n-3 ight )!!sqrt{2n-1}}{left ( 2n-2 ight )!!}cdot sqrt{frac{n}{2n-1}}\ &=frac{pi }{2}cdot sqrt{frac{2}{pi }}cdot frac{1}{sqrt{2}}=frac{sqrt{pi }}{2} end{align*}]
方法3:考虑两个含参变量积分
[fleft ( x ight )=left ( int_{0}^{x}e^{-t^{2}}mathrm{d}t ight )^{2}~~,~~gleft ( x ight )=int_{0}^{1}frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}}mathrm{d}u]
利用积分号下微分法,得
[egin{align*} f'left ( x ight )&= 2e^{-x^{2}}int_{0}^{x}e^{-t^{2}}mathrm{d}t\ g'left ( x ight )&=int_{0}^{1}frac{partial }{partial x}left [ frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}} ight ]mathrm{d}u=-2xe^{-x^{2}}int_{0}^{1}e^{-x^{2}u^{2}}mathrm{d}u end{align*}]
对后一积分,令(xu=t), 则
[g'left ( x ight )=-2xe^{-x^{2}}int_{0}^{x}e^{-t^{2}}mathrm{d}t=-f'left ( x ight )~~~left ( xgeq 0 ight )]
于是
[egin{align*} fleft ( x ight )+gleft ( x ight )=c~~~left ( xgeq 0 ight ) ag1 end{align*}]
由于 (displaystyle f(0)=0 , gleft ( 0 ight )=frac{pi }{4}), 故 (c=dfrac{pi }{4}), 即
[f(x)+g(x)=dfrac{pi }{4}~~~left ( xgeq 0 ight )]
当 (uin left [ 0,1 ight ]), 有
[0leq frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}}leq e^{-x^{2}u^{2}}leq e^{-x^{2}}~~~left ( xgeq 0 ight )]
因此,当 $x ightarrow infty $ 时,函数 (displaystyle frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}}) 关于 (uin left [ 0,1 ight ]) 一致的趋于0.
[lim_{x ightarrow infty }gleft ( x ight )=lim_{x ightarrow infty }int_{0}^{1}frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}}mathrm{d}u=int_{0}^{1}lim_{x ightarrow infty }frac{e^{-x^{2}left ( 1+u^{2} ight )}}{1+u^{2}}mathrm{d}u=0]
从而,由 (f(x)) 的定义及(1),得
[I=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=sqrt{lim_{x ightarrow infty }fleft ( x ight )}=sqrt{lim_{x ightarrow infty }frac{pi }{4}-gleft ( x ight )}=sqrt{frac{pi }{4}}=frac{sqrt{pi }}{2}]
方法4:设 (displaystyle fleft ( t ight )=int_{0}^{infty }e^{-tx^{2}}mathrm{d}x), 对 (f(t)) 取拉普拉斯变换,得
[mathcal{L}left ( int_{0}^{infty }e^{-tx^{2}}mathrm{d}x ight )=int_{0}^{infty }int_{0}^{infty }e^{-tx^{2}}e^{-st}mathrm{d}tmathrm{d}x=int_{0}^{infty }mathcal{L}left ( e^{-tx^{2}} ight )mathrm{d}x=int_{0}^{infty }frac{mathrm{d}x}{s+x^{2}}=frac{pi }{2sqrt{s}}]
再取拉普拉斯逆变换,有 (displaystyle fleft ( t ight )=int_{0}^{infty }e^{-tx^{2}}mathrm{d}x=frac{sqrt{pi }}{2sqrt{t}}), 在上式中,令 (t=1), 则
[I=fleft ( 1 ight )=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=frac{sqrt{pi }}{2}]
方法5:这种利用伽马函数的方法应该是高数中第一次接触的,出现在同济高数上册第五章最后,不过教材中打了星号,所以多数人都不了解,首先我们引入伽马函数的定义
[Gamma left ( x ight )=int_{0}^{infty }t^{x-1}e^{-t}mathrm{d}t]
所以,我们令 (x^2=t), 有
[I=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=frac{1}{2}int_{0}^{infty }t^{-frac{1}{2}}e^{-t}mathrm{d}t=frac{1}{2}Gamma left ( frac{1}{2} ight )=frac{sqrt{pi }}{2}]
其中 (displaystyle Gamma left ( frac{1}{2} ight )=sqrt{pi }) 可利用余元公式求得,这里不做证明.
方法6: 不难证明,函数 ((1+t)e^{-t}) 在 (t=0) 时达到它的最大值1.因此当 (t eq 0) 时,((1+t)e^{-t}<1), 令 (t=pm x^{2}), 即得
[left ( 1-x^{2} ight )e^{x^{2}}<1~,~left ( 1+x^{2} ight )e^{-x^{2}}<1]

[left ( 1-x^{2} ight )<e^{-x^{2}}<frac{1}{1+x^{2}}~~~left ( x>0 ight )]
假设限定第一个不等式中的 (x) 在(0,1)内变化,而第二个不等式中 (x) 则看作是任意的,把上式同 (n) 次方,有
[left ( 1-x^{2} ight )^{n}<e^{-nx^{2}}~~~left ( 0<x<1 ight )]
[e^{-nx^{2}}<frac{1}{left ( 1+x^{2} ight )^{n}}~~~left ( x>0 ight )]
第一个不等式即从0到1积分,第二个不等式取从0到(+infty)的积分,得
[int_{0}^{1}left ( 1-x^{2} ight )^{n}mathrm{d}x<int_{0}^{1}e^{-nx^{2}}mathrm{d}x<int_{0}^{infty }e^{-nx^{2}}mathrm{d}x<int_{0}^{infty }frac{1}{left ( 1+x^{2} ight )^{n}}mathrm{d}x]
在 (displaystyle int_{0}^{1}left ( 1-x^{2} ight )^{n}mathrm{d}x) 中,令 (x=cos t), 则
[int_{0}^{1}left ( 1-x^{2} ight )^{n}mathrm{d}x=int_{0}^{frac{pi }{2}}sin^{2n+1}tmathrm{d}t=frac{left ( 2n ight )!!}{left ( 2n+1 ight )!!}]
在 (displaystyle int_{0}^{infty }frac{1}{left ( 1+x^{2} ight )^{n}}mathrm{d}x) 中,令 (x=cot t), 则
[int_{0}^{infty }frac{1}{left ( 1+x^{2} ight )^{n}}mathrm{d}x=int_{0}^{frac{pi }{2}}sin^{2n-2}tmathrm{d}t=frac{left ( 2n-3 ight )!!}{left ( 2n-2 ight )!!}cdot frac{pi }{2}]
在 (displaystyle int_{0}^{infty }e^{-nx^{2}}mathrm{d}x) 中,令 (displaystyle x=frac{t}{sqrt{n}}), 则
[int_{0}^{infty }e^{-nx^{2}}mathrm{d}x=frac{1}{sqrt{n}} int_{0}^{infty }e^{-t^{2}}mathrm{d}t=frac{1}{sqrt{n}}I]
综上所述
[sqrt{n}cdot frac{left ( 2n ight )!!}{left ( 2n+1 ight )!!}<I<sqrt{n}cdot frac{left ( 2n-3 ight )!!}{left ( 2n-2 ight )!!}cdot frac{pi }{2}]
取平方得
[frac{n}{2n+1}cdot frac{left [ left ( 2n ight )!! ight ]^{2}}{left ( 2n+1 ight )left [ left ( 2n-1 ight )!! ight ]^{2}}<I^2<frac{n}{2n-1}cdot frac{left [ left ( 2n-3 ight )!! ight ]^{2}}{left [ left ( 2n-2 ight )!! ight ]^{2}}cdot left ( frac{pi }{2} ight )^{2}]
根据Wallis公式
[frac{pi }{2}=lim_{n ightarrow infty }frac{left [ left ( 2n ight )!! ight ]^{2}}{left ( 2n+1 ight )left [ left ( 2n-1 ight )!! ight ]^{2}}]
不等式两边当 $n ightarrow infty $ 时的极限都是 (dfrac{pi }{4}), 所以
[I^2=frac{pi }{4}Rightarrow I=frac{sqrt{pi }}{2}]
方法7:当然也可以利用三重积分
[egin{align*} &8I^3 = int_{-infty}^{infty} int_{-infty}^{infty} int_{-infty}^{infty}e^{-x^2 - y^2 - z^2}\,mathrm{d}x\,mathrm{d}y\,mathrm{d}z= int_{-infty}^{infty} int_{-infty}^{infty} int_{-infty}^{infty}e^{-x^2 - y^2 - z^2}\,mathrm{d}x\,mathrm{d}y\,mathrm{d}z\ Rightarrow &8I^3 = 4piint_0^{infty} ho^2 e^{- ho^2}\,mathrm{d} ho= 2pi int_0^{infty} e^{- ho^2}\,mathrm{d} ho=2pi cdot {I} Rightarrow 8I^3=2pi IRightarrow I=frac{sqrt{pi }}{2} end{align*}]
方法8:注意到
[n! =int_0^infty e^{-sqrt[n]x} mathrm{d}xifffrac1n! =int_0^infty e^{-x^n}mathrm{d}x ightarrowfrac12! = int_0^infty e^{-x^2}mathrm{d}x]
[int_0^1Big(1-sqrt[n]xBig)^m\,mathrm{d}x = int_0^1Big(1-sqrt[m]xBig)^n\,mathrm{d}x = frac1{C_{m+n}^n} = frac1{C_{m+n}^m} = frac{m!\,n!}{(m+n)!}]
所以我们有
[fracpi4 = int_0^1sqrt{1-x^2}\,mathrm{d}x = frac{left(dfrac12! ight)^2}{left(dfrac12 + dfrac12 ight)!} =left(frac12! ight)^2]
所以
[I=int_0^infty e^{-x^2}mathrm{d}x = frac12! = sqrt{piover4} = frac{sqrtpi}2]
方法9:利用
[int_0^infty e^{-x^2}mathrm{d}x=sqrt pi int_0^infty frac{1}{sqrt pi}e^{-x^2}mathrm{d}x=sqrt pi P(Xgeq0)]
其中
[Xsim Nleft ( 0,frac{1}{2} ight )~~,~~P(X>0)=P(X>E(X))=frac{1}{2}]
所以
[I=int_{0}^{infty }e^{-x^{2}}mathrm{d}x=frac{sqrt{pi }}{2}]
方法10:利用
[F(omega) = frac{1}{2 pi} int_{-infty}^{+infty} expleft(frac{-t^2}{2} ight) exp(- i omega t) mathrm{d}t]
所以
[F(omega) = frac{1}{pi} int_{0}^{+infty} expleft(frac{-t^2}{2} ight) cos( omega t) mathrm{d}t]
所以我们有
[F'(omega) = - omega F(omega)Rightarrow F(omega) = C expleft(frac{-omega^2}{2} ight)]

[expleft(frac{-x^2}{2} ight) = int_{-infty}^{+infty} F(omega) exp( i omega x) mathrm{d}omega]
可得 (displaystyle C = frac{1}{sqrt{2pi}}.) 令 (omega=0), 有
[F(0) = C = frac{1}{2 pi} int_{-infty}^{+infty} expleft(frac{-t^2}{2} ight) mathrm{d}t]
所以
[sqrt{2} int_{-infty}^{+infty} expleft(-t^2 ight) mathrm{d}t = sqrt{2pi}Rightarrow I=frac{sqrt{pi }}{2}.]

鬼斧神工:求n维球的体积 - Renascence_5

原文地址:http://spaces.ac.cn/archives/3154/
原文作者:苏剑林


标准思路
简单来说,(n)维球体积就是如下(n)重积分
[V_n(r)=int_{x_1^2+x_2^2+dots+x_n^2leq r^2}mathrm{d}x_1 mathrm{d}x_2dots mathrm{d}x_n]
用更加几何的思路,我们通过一组平行面((n−1)维的平行面)分割,使得n维球分解为一系列近似小柱体,因此,可以得到递推公式
[V_n (r)=int_{-r}^r V_{n-1} left(sqrt{r^2-t^2} ight)mathrm{d}t]
设(t=rsin heta_1),就有
[V_n (r)=rint_{-frac{pi}{2}}^{frac{pi}{2}} V_{n-1} left(rcos heta_1 ight)cos heta_1 mathrm{d} heta_1]
迭代一次就有
[V_n (r)=r^2int_{-frac{pi}{2}}^{frac{pi}{2}}int_{-frac{pi}{2}}^{frac{pi}{2}} V_{n-2} left(rcos heta_1cos heta_2 ight)cos heta_1cos^2 heta_2 mathrm{d} heta_1 mathrm{d} heta_2]
迭代(n−1)次
[egin{align*}V_n (r)=&r^{n-1}int_{-frac{pi}{2}}^{frac{pi}{2}}dotsint_{-frac{pi}{2}}^{frac{pi}{2}}int_{-frac{pi}{2}}^{frac{pi}{2}} V_1left(rcos heta_1cos heta_2dots cos heta_{n-1} ight) imes\ &cos heta_1cos^2 heta_2dotscos^{n-1} heta_{n-1} mathrm{d} heta_1 mathrm{d} heta_2dots mathrm{d} heta_{n-1}end{align*}]
其中(V_1(r)=2r),即两倍半径长的线段。从而
[V_n (r)=2r^{n}int_{-frac{pi}{2}}^{frac{pi}{2}}dotsint_{-frac{pi}{2}}^{frac{pi}{2}}int_{-frac{pi}{2}}^{frac{pi}{2}} cos^2 heta_1cos^3 heta_2dotscos^{n} heta_{n-1} mathrm{d} heta_1 mathrm{d} heta_2dots mathrm{d} heta_{n-1}]
完成这个积分,最终就得到n维球体积的公式,这个积分自然是可以求出来的(只是(n−1)个一维积分的乘积)。但是这样的步骤太不容易了,为了将其跟伽马函数联系起来,还要做很多工作。总的来说,这是一个不容易记忆、也不怎么漂亮的标准方法。


绝妙思路
有一个利用高斯积分的绝妙技巧,能够帮助我们直接将球体积跟伽马函数联系起来,整个过程堪称鬼斧神工,而且给人“仅此一家,别无分号”的感觉。据说这个技巧为物理系学生所知晓,我是从百读文库看到的,原始来源则是《热力学与统计力学》顾莱纳(德),例5.2 理想气体的熵的统计计算。

这一绝妙的思路,始于我们用两种不同的思路计算高斯积分
[egin{align*} G(n)=int_{-infty}^{+infty}dotsint_{-infty}^{+infty}int_{-infty}^{+infty} expleft(-x_1^2-x_2^2-dots-x_n^2 ight)mathrm{d}x_1 mathrm{d}x_2 dots mathrm{d}x_n ag{1} end{align*}]
一方面,将((1))当作(n)次累次积分,因为我们已经算得
[int_{-infty}^{+infty}exp(-t^2)mathrm{d}t=sqrt{pi}]
而((1))只不过是这样的(n)个积分的乘积,因此
[egin{align*} G(n)=pi^{n/2} ag{2} end{align*}]
另一方面,将((1))当作(n)重积分,由于积分变量只是跟径向长度(r=sqrt{x_1^2+x_2^2+dots+x_n^2})有关的变量,因此很容易联想到球坐标,在(n)维空间中,可以称为“超球坐标”,不需要将超球坐标完整写出来,只需要注意到,球内的积分,可以化为先对“球壳”进行积分,然后再对球半径进行积分。
[egin{align*} G(n)=int_{0}^{+infty}mathrm{d}rint_{S_n(r)}expleft(-r^2 ight)mathrm{d}S_n ag{3} end{align*}]
这里的(S_n(r))是半径为(r)的(n)维球体表面(以及表面积,在不至于混淆的情况下,这里不作区分)。但是注意到,被积函数只跟(r)有关,因此对球表面进行积分,等价于原函数乘以球的表面积而已,因此((2))式的结果为
[egin{align*} G(n)=int_{0}^{+infty}mathrm{d}rexpleft(-r^2 ight)S_n(r) ag{4} end{align*}]
虽然我们不知道(n)维球的体积和表面积公式,但是我们可以肯定,(n)维球的体积一定正比于(r^n),即有
[V_n (r)=V_n(1)r^n]
球的表面积,就是球体积的一阶导数(考虑球壳分割),那么
[S_n (r)=n V_n(1)r^{n-1}]
代入((4)),得到
[egin{align*}G(n)=&n V_n(1)int_{0}^{+infty}r^{n-1}expleft(-r^2 ight)mathrm{d}r\ =&frac{1}{2}n V_n(1)int_{0}^{+infty}(r^2)^{n/2-1}expleft(-r^2 ight)mathrm{d}(r^2)\ =&frac{1}{2}n V_n(1)int_{0}^{+infty}z^{n/2-1}expleft(-z ight)mathrm{d}zquadleft(z=r^2 ight)\ =&frac{1}{2}n V_n(1)Gammaleft(frac{n}{2} ight) ag{5}end{align*}]
结合((2))得
[pi^{n/2}=G(n)=frac{1}{2}n V_n(1)Gammaleft(frac{n}{2} ight)]
从而
[V_n(1)=frac{pi^{n/2}}{frac{1}{2}nGammaleft(dfrac{n}{2} ight)}=frac{pi^{n/2}}{Gammaleft(dfrac{n}{2}+1 ight)}]
最后
[Largeoxed{displaystyle V_n(r)=frac{pi^{n/2}}{Gammaleft(dfrac{n}{2}+1 ight)}r^n}]
就这样得到了(n)维球体积公式!!对(r)求导得到(n)维球表面积公式
[Largeoxed{displaystyle S_n(r)=frac{2pi^{n/2}}{Gammaleft(dfrac{n}{2} ight)}r^{n-1}}]
结合前后两个方法,就得到
[largeoxed{displaystyle color{red}{frac{pi^{n/2}}{Gammaleft(dfrac{n}{2}+1 ight)}=2int_{-frac{pi}{2}}^{frac{pi}{2}}dotsint_{-frac{pi}{2}}^{frac{pi}{2}}int_{-frac{pi}{2}}^{frac{pi}{2}} cos^2 heta_1cos^3 heta_2dotscos^{n} heta_{n-1} mathrm{d} heta_1 mathrm{d} heta_2dots mathrm{d} heta_{n-1}}}]

原文地址:https://www.cnblogs.com/Eufisky/p/9966979.html