当前位置:首页|资讯

泰勒公式间接展开法(拔高篇)

作者:现代微积分发布时间:2024-09-07

前言

话接上回。在前一篇文章

泰勒公式的证明和应用(基础篇)

中,我们推导出了泰勒公式,并且用直接法推出了几个常见函数的泰勒展开式,这里再附上当加深印象:

这种直接求n阶导再代值的方法称为“直接展开法”,直接展开法可用于展开若干个已知泰勒展开式的函数的线性组合。目前而言,我们可以展开如上几个函数的线性组合

如:y=eˣ+sinx-2ln(1+x),y=sinx+3cosx-5√(1+x)

这种线性组合的形式只需每项分别展开到需要的阶数即可

然而,我们也时常遇到初等函数相乘和复合的情形,例如:

y=eˣsinx,y=x*sin(x³)+ln(x²+1)

那么这时候直接求导就又比较复杂了,此时就得考虑“另辟蹊径”寻找一些间接展开的方法。这里就开门见山列举接下来要分享的3种方法

定理1:函数乘积的泰勒展开(用于解决两个乃至多个函数乘积的泰勒展开)

定理2:复合函数的泰勒展开(用于解决复合函数的泰勒展开)

定理3:导函数/变限积分函数的泰勒展开(用于解决某个函数的导函数/原函数的泰勒展开)

接下来逐一证明这几个定理并解决相关的练习题~(为叙述方便,下文提及定理1/2/3时就默认对应这几个定理了)

在此之前,有必要回顾一下无穷小量的一些运算性质(这些用同阶/高阶无穷小的定义即可证明)

%5Cbegin%7Balign%7D%0A%26x%5Cto%200%2C%5C%5C%0A%26%5Csum_%7Bi%3D1%7D%5E%7Bm%7D%20c_i%5Comicron%20(x%5E%5Calpha%20)%3D%5Comicron%20(x%5E%5Calpha%20)%5C%5C%0A%260%3C%5Calpha%3C%5Cbeta%20%5CRightarrow%20%0A%5Cleft%5C%7B%5Cbegin%7Bmatrix%7D%0Ac_1%20x%5E%5Calpha%20%2B%20c_2%20%20x%5E%5Cbeta%20%5Csim%20%20c_1%20x%5E%5Calpha%20%5C%5C%0Ac_1%5Comicron%20%20(x%5E%5Calpha%20)%2B%20c_2%5Comicron%20%20(x%5E%5Cbeta%20)%3D%20%5Comicron%20%20(x%5E%5Calpha%20)%0A%5Cend%7Bmatrix%7D%5Cright.%5C%5C%0A%26x%5E%5Cbeta%20%5Comicron%20(x%5E%5Calpha%20)%3D%5Comicron%20(x%5E%7B%5Calpha%20%2B%5Cbeta%20%7D)~%7B%5Ccolor%7BGray%7D%20%7B(%5Cbeta%20%5Cgeqslant-%5Calpha%20)%7D%7D%20%5C%5C%0A%5Cend%7Balign%7D

第一条是线性性质:比某个次幂高阶的无穷小的线性组合仍为该次幂的高阶无穷小;

第二条是无穷小的“短板效应”,对于两个不同阶的函数相加,高阶+低阶等价于低阶

第三条是幂函数与无穷小相乘的性质。

这里要提及一下第三条,这在“形式上”具有乘法性质:

看似是幂函数直接与o()中()里面的幂次相乘得x%5E%5Cbeta%20x%5E%7B%5Calpha%20%7D%3Dx%5E%7B%5Calpha%20%2B%5Cbeta%20%7D但实际上并不是这样运算高阶无穷小o(f(x))中的f(x)并不能赋值也不能与外界作运算,o(f(x))这整个式子的含义仅仅是比f(x)高阶的无穷小量(某个函数群体中的其中一个函数)

比如当x→0时,x²,2x²,-100x²,sin²x,tan²x,(tanx+sinx)²,x³等等这些函数都可以写成o(x)(即比x高阶的无穷小量),因此o(f(x))这些项,只是穿着某件外衣的一个函数罢了(这件外衣说明了它的一个性质:比f(x)高阶)

所以第三个等式其实是先证明分子与分母比值的极限为0再由高阶无穷小定义得出的:

%5Clim_%7Bx%20%5Cto%200%7D%20%5Cfrac%7Bx%5E%5Cbeta%20%5Comicron%20(x%5E%7B%5Calpha%20%7D)%7D%7Bx%5E%7B%5Calpha%20%2B%5Cbeta%20%7D%7D%20%0A%3D%5Clim_%7Bx%20%5Cto%200%7D%20%5Cfrac%7B%5Comicron%20(x%5E%7B%5Calpha%20%7D)%7D%7Bx%5E%7B%5Calpha%7D%7D%20%0A%3D0%0A%5CRightarrow%20x%5E%5Cbeta%20%5Comicron%20(x%5E%7B%5Calpha%20%7D)%3D%5Comicron%20%20(x%5E%7B%5Calpha%20%2B%5Cbeta%20%7D)

定理1:函数乘积的泰勒展开

若函数f(x)和g(x)满足:

x→0,f(x)=O(x^α),g(x)=O(x^β)

并且f(x)和g(x)在x=0处有充足的高阶导数。

若要将f(x)g(x)在x=0处展开到n阶,则f(x)需展开到n-β阶g(x)需展开到n-α阶

简单来说就是f(x)和g(x)可以展开足够多项,并且f(x),g(x)的最低阶项分别为x^α,x^β。那么要将f(x)g(x)整体展开到n阶,对于f(x)而言展开的阶数取决于g(x)的最低阶

g(x)而言展开的阶数取决于f(x)的最低阶

另外,上述定理的展开点选择在x=0处,在其他点的展开类比即可

证明:设f(x),g(x)分别展开到m₁,m₂阶,则有:

%5Cbegin%7Balign%7D%0A%26f(x)g(x)%5C%5C%0A%3D%26(a_%5Calpha%20x%5E%5Calpha%20%2Ba_%7B%5Calpha%20%2B1%7D%20x%5E%7B%5Calpha%2B1%7D%2B%5Ccdots%20%2Ba_%7Bm_1%7Dx%5E%7Bm_1%7D%2B%5Comicron%20(x%5E%7Bm_1%7D))%5Ccdot%5C%5C%0A%26(b_%5Cbeta%20x%5E%5Cbeta%20%2Bb_%7B%5Cbeta%20%2B1%7D%20x%5E%7B%5Cbeta%2B1%7D%2B%5Ccdots%20%2Bb_%7Bm_2%7Dx%5E%7Bm_2%7D%2B%5Comicron%20(x%5E%7Bm_2%7D))%5C%5C%0A%3D%26(p_1(x)%2B%5Comicron%20(x%5E%7Bm_1%7D))(p_2(x)%2B%5Comicron%20(x%5E%7Bm_2%7D))%5C%5C%0A%3D%26p_1(x)p_2(x)%2Bp_1(x)%5Comicron%20%20(x%5E%7Bm_2%7D)%2Bp_2(x)%5Comicron%20(x%5E%7Bm_1%7D)%2B%5Comicron%20(x%5E%7Bm_1%7D)%5Comicron%20(x%5E%7Bm_2%7D)%0A%5Cend%7Balign%7D

其中p₁(x),p₂(x)分别为f(x)的m₁阶泰勒多项式,g(x)的m₂阶泰勒多项式

接下来我们要把最后一行结果中带余项的部分进行化简:

其中

%5Cbegin%7Balign%7D%0A%26p_1(x)%5Comicron%20%20(x%5E%7Bm_2%7D)%5C%5C%0A%3D%26(a_%5Calpha%20x%5E%5Calpha%20%2B%5Comicron%20(x%5E%5Calpha%20))%5Comicron%20%20(x%5E%7Bm_2%7D)%5C%5C%0A%3D%26a_%5Calpha%20x%5E%5Calpha%20%5Comicron%20%20(x%5E%7Bm_2%7D)%2B%5Comicron%20(x%5E%5Calpha%20)%5Comicron%20%20(x%5E%7Bm_2%7D)%5C%5C%0A%3D%26a_%7B%5Calpha%20%7D%5Comicron%20%20(x%5E%7Bm_2%2B%5Calpha%20%7D)%2B%5Comicron%20(x%5E%7Bm_2%2B%5Calpha%20%7D%20)%5C%5C%0A%3D%26%5Comicron%20(x%5E%7Bm_2%2B%5Calpha%20%7D%20)%0A%5Cend%7Balign%7D


%5Cbegin%7Balign%7D%0A%26p_2(x)%5Comicron%20%20(x%5E%7Bm_1%7D)%5C%5C%0A%3D%26(b_%5Cbeta%20%20x%5E%5Cbeta%20%2B%5Comicron%20(x%5E%5Cbeta%20))%5Comicron%20%20(x%5E%7Bm_1%7D)%5C%5C%0A%3D%26b_%5Cbeta%20%20x%5E%5Cbeta%20%5Comicron%20%20(x%5E%7Bm_1%7D)%2B%5Comicron%20(x%5E%5Cbeta%20%20)%5Comicron%20%20(x%5E%7Bm_1%7D)%5C%5C%0A%3D%26b_%5Cbeta%20%5Comicron%20%20(x%5E%7Bm_1%2B%5Cbeta%20%7D)%2B%5Comicron%20(x%5E%7Bm_1%2B%5Cbeta%20%7D%20)%5C%5C%0A%3D%26%5Comicron%20(x%5E%7Bm_1%2B%5Cbeta%20%7D)%0A%5Cend%7Balign%7D


o(xᵐ¹)o(ᵐ²)=o(xᵐ¹⁺ᵐ²)


即得:

f(x)%3Dp_1(x)p_2(x)%2B%5Comicron%20(x%5E%7Bm_2%2B%5Calpha%20%7D%20)%2B%5Comicron%20(x%5E%7Bm_1%2B%5Cbeta%20%7D)%2B%5Comicron%20(x%5E%7Bm_1%2Bm_2%20%7D)%0A

要让f(x)展开到n阶,那么余项就得是o(xⁿ),因此根据无穷小的“短板效应”的性质(低阶+高阶~低阶)得:上式中的3个余项都得≥n阶,即有:

%5Cleft%5C%7B%5Cbegin%7Bmatrix%7D%0Am_2%2B%5Calpha%20%5Cgeqslant%20n%20%5C%5C%0Am_1%2B%5Cbeta%20%5Cgeqslant%20n%20%5C%5C%0Am_1%2Bm_2%5Cgeqslant%20n%0A%5Cend%7Bmatrix%7D%5Cright.%0A%5CRightarrow%20%0A%5Cleft%5C%7B%5Cbegin%7Bmatrix%7D%0Am_2%5Cgeqslant%20n%20-%5Calpha%20%5C%5C%0Am_1%5Cgeqslant%20n%20-%5Cbeta%20%0A%5Cend%7Bmatrix%7D%5Cright.%0A


例(1):将f(x)=eˣsinx在x=0处展开到3阶

先找各自的最低阶:

eˣ最低阶为0(与xº=1同阶);sinx最低阶为1(与x同阶);

由定理1,对于eˣ而言,需要展开的阶数为3-1=2;

对sinx而言,需要展开的阶数为3-0=3

也就是用需要展开的阶数n去减另外一项最低阶数

从而有:

%5Cbegin%7Balign%7D%0A%26%5Cmathrm%7Be%7D%20%5Ex%5Csin%20x%5C%5C%0A%3D%26(1%2Bx%2B%5Cfrac%7Bx%5E2%7D%7B2!%7D%2B%5Comicron%20(x%5E2)%20)(x-%5Cfrac%7Bx%5E3%7D%7B3!%7D%2B%5Comicron%20(x%5E3)%20)%5C%5C%0A%3D%26(1%2Bx%2B%5Cfrac%7Bx%5E2%7D%7B2%7D)(x-%5Cfrac%7Bx%5E3%7D%7B6%7D%20)%2B%5Comicron%20(x%5E3)%0A%5Cend%7Balign%7D

用上述准则选取阶数后,可以放心大胆地再后面写个o(x³)了,证明过程前文已经详细论述了

接下来继续将多项式相乘进行展开。这时仍不需要一次性全部展开,而是在展开过程中遇到比x³高次的可以直接扔进“回收站”(o(x³))里,后续过程如下:

%5Cbegin%7Balign%7D%0A%26(1%2Bx%2B%5Cfrac%7Bx%5E2%7D%7B2%7D)(x-%5Cfrac%7Bx%5E3%7D%7B6%7D%20)%2B%5Comicron%20(x%5E3)%5C%5C%0A%3D%26x%2Bx%5E2%2B%5Cfrac%7Bx%5E3%7D%7B2%7D%2B%5B(-%5Cfrac%7Bx%5E3%7D%7B6%7D%20)%2B%5Comicron%20%20(x%5E3)%20%5D%2B%5Comicron%20(x%5E3)%5C%5C%0A%3D%26x%2Bx%5E2%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Comicron%20(x%5E3)%0A%5Cend%7Balign%7D

第二行拿(-x³/6)与(1+x+x²/2)相乘时,乘到第一项之后,后面的项次数已经超过3了,因此后面的两项可以直接扔进“回收站”o(x³)里

%5Cmathrm%7Be%7D%20%5Ex%3Dx%2Bx%5E2%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Comicron%20(x%5E3)

我们再来看看展开不够会有何后果?

比如第一项只展开到了1阶:

%5Cmathrm%7Be%7D%20%5Ex%5Csin%20x%0A%3D(1%2Bx%2B%5Comicron%20(x)%20)(x-%5Cfrac%7Bx%5E3%7D%7B3!%7D%2B%5Comicron%20(x%5E3)%20)

我们知道,两个多项式相乘,就相当于从这两个多项式中分别选出一项进行相乘,最后把所有的结果相加得到最终的展开式

因此如果第一个括号选o(x),第二个括号选x,那么就可以组合出展开式中的其中一项:

x*o(x)=o(x²),这一项只说明了比x²高阶,如果这一项刚好与x³同阶,那这一项的系数就被遗漏掉了,简称“漏阶”。

说白了就是对于eˣ而言,我要保证余项与另一个因式(sinx)的最低阶相乘后都得是o(x³)才行,这也就解释了为什么要用n减去另外一项的最低阶的底层逻辑了。

对于sinx而言也是同理,sinx的余项与另一个因式(eˣ)的最低阶相乘后都得是o()才行。

对于定理1,由数学归纳法可以推广到多个函数相乘的泰勒展开:

定理1拓展:多个函数乘积的泰勒展开

若函数f₁(x),f₂(x),...,fₘ(x)满足:

x%5Cto%200%2Cf_1(x)%3DO(x%5E%7B%5Calpha_1%7D%20)%20%2Cf_2(x)%3DO%20(x%5E%7B%5Calpha_2%7D)%2C%5Ccdots%20%2Cf_m(x)%3DO(x%5E%7B%5Calpha_m%7D)

并且f₁(x),f₂(x),...,fₘ(x)在x=0处有充足的高阶导数。

若要将f₁(x),f₂(x),...,fₘ(x)在x=0处展开到n阶,则有:

f₁(x)需展开到n-(a₂+a₃+...+aₘ)阶;

f₂(x)需展开到n-(a₁+a₃+...+aₘ)阶;

f₃(x)需展开到n-(a₁+a₂+a₄+...+aₘ)阶;

...;

fₘ(x)需展开到n-(a₂+a₃+...+aₘ₋₁)阶


也就是其中一项需要展开的阶数,等于n减去其余因式的最低阶数之和

例(2):将f(x)=eˣsinx√(1+x)在x=0处展开到3阶

分别求出各个因子的阶数:

eˣ阶数为0(与xº同阶);sinx阶数为1(与x¹同阶);√(1+x)阶数为0(与xº同阶)

由定理1(拓展)得:

对于eˣ而言,需要展开的阶数为:3-(1+0)=2;

对于sinx而言,需要展开的阶数为:3-(0+0)=3;

对于√(1+x)而言,需要展开的阶数为:3-(0+1)=2

在确定各自需要展开的阶数后,剩下的就是暴力展开合并同类项了~

%5Cbegin%7Balign%7D%0Af(x)%26%3D%5Cmathrm%7Be%7D%5Ex%5Csin%20x%5Csqrt%7B1%2Bx%7D%5C%5C%0A%26%3D(1%2Bx%2B%5Cfrac%7Bx%5E2%7D%7B2!%7D%20%2B%5Comicron%20(x%5E3))%20(x-%5Cfrac%7Bx%5E3%7D%7B3!%7D%2B%5Comicron%20%20(x%5E3)%20)%5C%5C%0A%26%5Ccdot%20(1%2B%5Cfrac%7B1%7D%7B2%7Dx%2B%5Cfrac%7B%5Cfrac%7B1%7D%7B2%7D(%5Cfrac%7B1%7D%7B2%7D-1%20)%20%7D%7B2%7Dx%5E2%2B%20%5Comicron%20%20(x%5E2)%20%20)%5C%5C%0A%26%3D(1%2Bx%2B%5Cfrac%7Bx%5E2%7D%7B2%7D%20)(x-%5Cfrac%7Bx%5E3%7D%7B6%7D%20)(1%2B%5Cfrac%7B1%7D%7B2%7Dx-%5Cfrac%7B1%7D%7B8%7D%20x%5E2)%0A%2B%5Comicron%20(x%5E3)%5C%5C%0A%26%3D(x%2B%5Cfrac%7Bx%5E2%7D%7B2%7D%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Comicron%20(x%5E3)%20%20)(1%2B%5Cfrac%7B1%7D%7B2%7Dx-%5Cfrac%7B1%7D%7B8%7D%20x%5E2)%0A%2B%5Comicron%20(x%5E3)%5C%5C%0A%26%3D(x%2B%5Cfrac%7Bx%5E2%7D%7B2%7D%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%20)(1%2B%5Cfrac%7B1%7D%7B2%7Dx-%5Cfrac%7B1%7D%7B8%7D%20x%5E2)%0A%2B%5Comicron%20(x%5E3)%5C%5C%0A%26%3Dx%2B%5Cfrac%7B3%7D%7B2%7D%20x%5E2%2B%5Cfrac%7B17%7D%7B24%7Dx%5E3%20%2B%5Comicron%20(x%5E3)%0A%5Cend%7Balign%7D


ps:在展开化简的过程中,也可以遵循定理1的展开准则,比如上式第3个等号开始,要将(1+x+x²/2)(x-x³/6)展开,展开到的阶数取决于其余项的最低阶(此处即1+x/2-x²/8的最低阶为0),因此(1+x+x²/2)(x-x³/6)展开到3-0=3阶就足够了,后面的高次项就可以扔进“回收站”里


以下练习题留作思考,感兴趣的读者可以先自行尝试:

f(x)%3D%5Cprod_%7Bk%3D1%7D%5E%7Bn%7D%20%5Csin%20kx%20%3D%5Csin%20x%5Csin%202x%5Ccdots%20%5Csin%20nx在x=0处展开到n+2阶

%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20k%5E2%3D%5Cfrac%7Bn(n%2B1)(2n%2B1)%7D%7B6%7D%20


f(x)%3Dn!x%5En-%5Cfrac%7Bn!n(n%2B1)(2n%2B1)%7D%7B36%7D%20x%5E%7Bn%2B2%7D%2B%5Comicron%20(x%5E%7Bn%2B2%7D)


定理2:复合函数泰勒展开

已知函数f(x)在x=x₀处具有n阶导数。且当t→t₀时函数g(t)→x₀,则复合函数f(g(t))在t=t₀处可展开为:

%20%5Cbegin%7Balign%7D%0Af(g(t))%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(g(t)-x_0)%5Ek%20%20%2B%5Comicron%20%20((g(t)-x_0)%5En)%5C%5C%0A%26%3Df(x_0)%2Bf'(x_0)(g(t)-x_0)%2B%5Ccdots%20%5C%5C%0A%26%2B%5Cfrac%7Bf%5E%7B(n)%7D(x_0)%7D%7Bn!%7D(g(t)-x_0)%5En%2B%5Comicron%20%20((g(t)-x_0)%5En)%0A%5Cend%7Balign%7D


证明:由f(x)在x=x₀处具有n阶导数,进而可将f(x)在x=x₀处展开为:

f(x)%3D%5Cunderbrace%7B%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%7D_%7BT_n(x)%7D%20%20%2B%0A%5Cunderbrace%7B%5Comicron%20%20((x-x_0)%5En)%7D_%7BR_n(x)%7D

%5Clim_%7Bx%20%5Cto%20x_0%7D%5Cfrac%7BR_n(x)%7D%7B(x-x_0)%5En%7D%20%3D%20%5Clim_%7Bx%20%5Cto%20x_0%7D%5Cfrac%7Bf(x)-T_n(x)%7D%7B(x-x_0)%5En%7D%3D0

以上是前一篇文章 泰勒公式的证明和应用(基础篇) 开头证明过的定理哦


又由于当t→t₀时g(t)→x₀,于是由极限的第一复合法则得:

%5Clim_%7Bt%20%5Cto%20t_0%7D%5Cfrac%7BR_n(g(t))%7D%7B(g(t)-x_0)%5En%7D%20%3D%20%5Clim_%7Bt%20%5Cto%20t_0%7D%5Cfrac%7Bf(g(t))-T_n(g(t))%7D%7B(g(t)-x_0)%5En%7D%3D0

进而由高阶无穷小的定义得:

t→t₀,f(g(t))-Tₙ(g(t))=o((g(t)-t₀)ⁿ)

也即

%5Cbegin%7Balign%7D%0Af(g(t))%26%3DT_n(g(t))%2B%5Comicron%20((g(t)-t_0)%5En)%5C%5C%0A%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(g(t)-x_0)%5Ek%20%20%2B%0A%5Comicron%20%20((g(t)-x_0)%5En)%0A%5Cend%7Balign%7D

定理得证!

为了方便记忆上述定理,可考虑将g(t)换成方框"□",即:

%5Cbbox%5B%238ffa9a%5D%7B%5CBox%20%5Cto%20x_0%2Cf(%5CBox)%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(%5CBox-x_0)%5Ek%20%20%2B%5Comicron%20%20((%5CBox-x_0)%5En)%7D

例(3):将函数f(x)=exp(x²)展开到4阶

外层函数为指数函数,于是由eˣ的泰勒公式以及定理2得:

%5CBox%20%5Cto%200%2C%5Cmathrm%7Be%7D%20%5E%7B%5CBox%7D%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7B%5CBox%5Ek%7D%7Bk!%7D%20%20%2B%5Comicron%20%20(%5CBox%5En)

余项的阶数取决于方框□的阶数,此处□=x²,因此□的阶数为2,进而□ⁿ=x²ⁿ的阶数为2n

要展开到4阶,则需2n≥4,即n≥2

取n=2,□=x²即得:

%5Cbegin%7Balign%7D%0A%5Cmathrm%7Be%7D%20%5E%7Bx%5E2%7D%26%3D1%2B(x%5E2)%2B%5Cfrac%7B(x%5E2)%5E2%7D%7B2!%7D%20%2B%5Comicron%20%20((x%5E2)%5E2)%5C%5C%0A%26%3D1%2Bx%5E2%2B%5Cfrac%7B1%7D%7B2%7D%20x%5E4%2B%5Comicron%20%20(x%5E4)%0A%5Cend%7Balign%7D


例(4):将函数f(x)=cos(sinx)展开到4阶

外层函数为余弦函数,于是由cosx的泰勒公式以及定理2得:

%5CBox%20%5Cto%200%2C%5Ccos%20%5CBox%20%3D1-%5Cfrac%7B%5CBox%20%5E2%7D%7B2!%7D%2B%5Ccdots%20%2B%5Cfrac%7B(-1)%5En%7D%7B(2n)!%7D%5CBox%20%5E%7B2n%7D%2B%5Comicron%20(%5CBox%20%5E%7B2n%7D)

ps:由于cosx的泰勒展开式中奇次项全为0,因此展开到第2n项和展开到第2n+1项效果是一样的,因此上述公式中的余项也可以写成o(□²ⁿ⁺¹)。一般而言要展开到奇数阶就选2n+1,展开到偶数阶就选2n,比如这题就选择展到第2n项

余项的阶数取决于方框□的阶数,此处□=sinx,因此□的阶数为1(与x同阶),进而□²ⁿ=(sinx)²ⁿ~x²ⁿ,即阶数为2n

要展开到4阶,则需2n≥4,即n≥2。

取n=2,□=sinx即得:

%5Cbegin%7Balign%7D%0A%5Ccos(%5Csin%20x)%26%20%3D1-%5Cfrac%7B(%5Csin%20x)%5E2%7D%7B2!%7D%2B%5Cfrac%7B(%5Csin%20x)%5E4%7D%7B4!%7D%2B%5Comicron%20((%5Csin%20x)%5E4)%5C%5C%0A%26%20%3D1-%5Cfrac%7B(%5Csin%20x)%5E2%7D%7B2%7D%2B%5Cfrac%7B(%5Csin%20x)%5E4%7D%7B24%7D%2B%5Comicron%20(x%5E4)%0A%5Cend%7Balign%7D

-%5Cfrac%7B(%5Csin%20x)%5E2%7D%7B2%7D%2C%5Cfrac%7B(%5Csin%20x)%5E4%7D%7B24%7D这两项各自都需要展开到4阶。而这是函数乘积的展开,这时就得用到定理1了。

sin²x=sinx*sinx

sinx的阶数为1(与x同阶)

则对于第一个sinx,需要展开的阶数为:4-1=3

ps:这里的减数1是指另一个sinx的阶数

同理对于第二个sinx,也需展开到4-1=3阶

于是有:

%5Cbegin%7Balign%7D%0A%5Csin%20%5E2x%26%3D(x-%5Cfrac%7Bx%5E3%7D%7B3!%7D%2B%5Comicron%20%20(x%5E3)%20)%5E2%5C%5C%0A%26%3D(x-%5Cfrac%7Bx%5E3%7D%7B3!%7D)%5E2%2B%5Comicron%20%20(x%5E4)%20%5C%5C%0A%26%3Dx%5E2-%5Cfrac%7Bx%5E4%7D%7B3%7D%20%2B%5Comicron%20%20(x%5E4)%20%0A%5Cend%7Balign%7D


对sin⁴x展开同理:

sinx的阶数为1(与x同阶)

则对于第一个sinx,需要展开的阶数为:4-(1+1+1)=1

同理对于其余的sinx,也需展开到4-(1+1+1)=1阶

于是有:

%5Csin%20%5E4x%3D(x%2B%5Comicron%20%20(x)%20)%5E4%3Dx%5E4%2B%5Comicron%20%20(x%5E4)%20


回代原式,最终整理得:

%5Cbegin%7Balign%7D%0A%5Cmathrm%7BLHS%7D%20%26%20%3D1-%5Cfrac%7B(%5Csin%20x)%5E2%7D%7B2%7D%2B%5Cfrac%7B(%5Csin%20x)%5E4%7D%7B24%7D%2B%5Comicron%20(x%5E4)%5C%5C%0A%26%3D1-%5Cfrac%7Bx%5E2-%5Cfrac%7Bx%5E4%7D%7B3%7D%20%2B%5Comicron%20%20(x%5E4)%20%7D%7B2%7D%20%2B%5Cfrac%7Bx%5E4%2B%5Comicron%20%20(x%5E4)%20%7D%7B24%7D%20%2B%5Comicron%20(x%5E4)%5C%5C%0A%26%3D1-%5Cfrac%7Bx%5E2%7D%7B2%7D%2B%5Cfrac%7B5x%5E4%7D%7B24%7D%20%20%2B%5Comicron%20(x%5E4)%0A%5Cend%7Balign%7D


可见,对于一般的复合函数进行泰勒展开,需要先用定理2展开外层(展开阶数取决于内层阶数的n倍,即□ⁿ的阶数)再用定理1逐项展开内层,二者需要搭配使用。因此对于复合函数的展开,可谓是泰勒展开中的硬骨头了。


定理3:导函数/原函数的泰勒展开

已知函数f(x)在x=x₀处具有n阶导数。则函数f(x)可展开为:

f(x)%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%20%2B%5Comicron%20%20((x-x_0)%5En)

其导函数/变限积分函数可展开为:

f'(x)%3D%20%5Cfrac%7B%5Cmathrm%7Bd%7D%20%7D%7B%5Cmathrm%7Bd%7D%20x%7D%5B%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%20%5D%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)

%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%20%3D%5Cint_%7Bx_0%7D%5E%7Bx%7D%20%5B%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(t-x_0)%5Ek%20%20%5D%5Cmathrm%7Bd%7Dt%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)

也即在“形式上”对展开式进行了求导or积分(把右边所有关于x的多项式进行求导or积分)。

当然这只是个辅助记忆的方法。前面已经说了o(f(x))中里面的f(x)并不能赋值和参与外部的任何运算,整个“o(f(x))”这个式子仅仅代表比f(x)高阶的无穷小量,它只是某个群体中的一员用该群体的总名称替代了而已。

x%5E2%2C2x%5E2%2C100x%5E2%2C-999x%5E2%2C%5Csin%5E2%20x%2C%5Ctan%20%5E22x%2C(%5Csin%20x%2B%5Csin%202x)%5E2%2C%5Ccdots这些函数都可以写成O(x²)(与x同阶的无穷小量)或o(x)(比x高阶的无穷小量)

定理3的具体的证明如下:

f(x)在x=x₀处具有n阶导数,则函数f(x)可展开为:

f(x)%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%20%2B%5Comicron%20%20((x-x_0)%5En)

这个就是前篇文章推导过的泰勒公式了

Case[1]:那么对于导函数f'(x)而言,就有:f'(x)在x=x₀处具有n-1阶导数

这是由于f(x)和f'(x)只“相差了一代”,f(x)有n代(n阶导)那么f'(x)就有n-1代(n-1阶导)

记g(x)=f'(x),对函数g(x)可展开为:

g(x)%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn-1%7D%5Cfrac%7Bg%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5E%7Bk%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)

又由g(x)=f'(x)得:g⁽ᵏ⁾(x₀)=f⁽ᵏ⁺¹⁾(x₀)

即g(x₀)=f'(x₀),g'(x₀)=f''(x₀),g''(x₀)=f'''(x₀),以此类推

f'(x)%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn-1%7D%5Cfrac%7Bf%5E%7B(k%2B1)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5E%7Bk%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)

其中k是从0求和到n-1,那么k+1就是从1求和到n,于是对求和变量作换元i=k+1得:

f'(x)%3D%20%5Csum_%7Bi%3D1%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(i)%7D(x_0)%7D%7B(i-1)!%7D(x-x_0)%5E%7Bi-1%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)

又由于求和式与求和变量的记号无关,因此将i再换成k即得:

f'(x)%3D%20%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7B(k-1)!%7D(x-x_0)%5E%7Bk-1%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)

则定理的前半部分得证!

ps:如果对于求和变量换元有点蒙的,可以先自己理解一下以下求和式的代数意义:

%5Csum_%7Bk%3D0%7D%5E%7B10%7D%202%5Ek%3D%5Csum_%7Bi%3D1%7D%5E%7B11%7D%202%5E%7Bi-1%7D%3D%5Csum_%7Bk%3D1%7D%5E%7B11%7D%202%5E%7Bk-1%7D

这三者表示的本质上都是同一个式子:2º+2¹+...+2¹º


对比两式:

%5Cleft%5C%7B%5Cbegin%7Bmatrix%7D%0A%5Cdisplaystyle%20%20f(x)%3D%20%7B%5Ccolor%7BBlue%7D%20%7B%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%7D%7D%20%20%2B%5Comicron%20%20((x-x_0)%5En)%5C%5C%0A%5Cdisplaystyle%20%20f'(x)%3D%20%7B%5Ccolor%7BBlue%7D%20%7B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7B(k-1)!%7D(x-x_0)%5E%7Bk-1%7D%20%7D%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn-1%7D)%0A%5Cend%7Bmatrix%7D%5Cright.

发现标蓝的部分,下者恰好可以由上者求导得到。且余项恰好也降低了一阶,因此形式上可以记忆为“两边求导”

但是还是得再强调一下,“两边求导”的口诀只是方便我们从形式上记忆该定理,实际上后者的o((x-x₀)ⁿ⁻¹)是对f'(x)整体展开得到的

对于变限积分也是同理

F(x)%3D%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt,满足F(x₀)=0,F'(x)=f(x)

Case[2]:f(x)在x=x₀处具有n阶导数,那么对于F(x)而言,就有:F(x)在x=x₀处具有n+1阶导数

这是由于f(x)和F(x)也只“相差了一代”,f(x)有n代(n阶导)那么F(x)就有n+1代(n+1阶导)

对函数F(x)可展开为:

F(x)%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn%2B1%7D%5Cfrac%7BF%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5E%7Bk%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)

又由F'(x)=f(x)得:F⁽ᵏ⁾(x₀)=f⁽ᵏ⁻¹⁾(x₀),k≥1

即F'(x₀)=f(x₀),F''(x0)=f'(x₀),F'''(x0)=f''(x₀),以此类推

%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%3D%20%5Cunderbrace%7B%200%7D_%7BF(x_0)%3D0%7D%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%2B1%7D%5Cfrac%7Bf%5E%7B(k-1)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5E%7Bk%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)

其中k是从1求和到n+1,那么k-1就是从0求和到n,于是对求和变换作换元i=k-1得:

%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%3D%20%5Csum_%7Bi%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(i)%7D(x_0)%7D%7B(i%2B1)!%7D(x-x_0)%5E%7Bi%2B1%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)

又由于求和式与求和变量的记号无关,因此将i再换成k即得:

%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%3D%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7B(k%2B1)!%7D(x-x_0)%5E%7Bk%2B1%7D%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)

则定理的后半部分得证!


对比两式:

%5Cleft%5C%7B%5Cbegin%7Bmatrix%7D%0A%5Cdisplaystyle%20f(x)%3D%20%7B%5Ccolor%7BGreen%7D%20%7B%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%7D%7D%20%20%2B%5Comicron%20%20((x-x_0)%5En)%20%5C%5C%0A%5Cdisplaystyle%20%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%3D%20%7B%5Ccolor%7BGreen%7D%20%7B%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7B(k%2B1)!%7D(x-x_0)%5E%7Bk%2B1%7D%20%7D%7D%20%2B%5Comicron%20%20((x-x_0)%5E%7Bn%2B1%7D)%0A%5Cend%7Bmatrix%7D%5Cright.

发现标绿的部分,下者恰好可以由上者积分得到。且余项恰好也升高了一阶,因此形式上可以记忆为“两边积分”

同理,“两边积分”的口诀也只是方便我们从形式上记忆该定理,实际上后者的o((x-x₀)ⁿ⁺¹)是对F(x)整体展开得到的


例(5):将f(x)%3D%5Cint_%7B0%7D%5E%7Bx%7D%20%5Cmathrm%7Be%7D%20%5E%7Bt%5E2%7D%5Cmathrm%7Bd%7Dt%20在x=0处展开到5阶

初步的想法是把原函数给积出来,再按照定理1、定理2进行展开。但很可惜这个被积函数已经被证明过不能用初等函数表示的,因此把原函数积出来这条路就行不通了。

而“积不出来”指的是“不能用初等函数”表示,但这个的原函数是一定存在的(由于连续函数必有原函数,并且变上限积分就是其中一个原函数),因此不妨碍我们对原函数进行展开。

由定理3,要对f(x)展开到5阶,只需对其“下一代”(此处即exp(x²))展开到5-1=4阶

%20%5Cmathrm%7Be%7D%20%5E%7Bx%5E2%7D前面我们刚好已经展开过了,用定理2展开得:

%5Cbegin%7Balign%7D%0A%5Cmathrm%7Be%7D%20%5E%7Bx%5E2%7D%26%3D1%2B(x%5E2)%2B%5Cfrac%7B(x%5E2)%5E2%7D%7B2!%7D%20%2B%5Comicron%20((x%5E2)%5E2)%5C%5C%0A%26%3D1%2Bx%5E2%2B%5Cfrac%7Bx%5E4%7D%7B2%7D%20%2B%5Comicron%20(x%5E4)%0A%5Cend%7Balign%7D

再由定理3得:

%5Cbegin%7Balign%7D%0A%5Cint_%7B0%7D%5E%7Bx%7D%20%5Cmathrm%7Be%7D%20%5E%7Bt%5E2%7D%5Cmathrm%7Bd%7Dt%20%26%3D%5Cint_%7B0%7D%5E%7Bx%7D%20(1%2Bt%5E2%2B%5Cfrac%7Bt%5E4%7D%7B2%7D%20)%5Cmathrm%7Bd%7Dt%2B%5Comicron%20(x%5E5)%5C%5C%0A%26%3Dx%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Cfrac%7Bx%5E5%7D%7B10%7D%20%2B%5Comicron%20(x%5E5)%0A%5Cend%7Balign%7D


其他应用(间接法推导几个常用的泰勒展开)

有了定理3,结合定理2,我们就能在现有的几个泰勒公式中就又可以推导出其他函数的泰勒展开式啦!

前篇文章已经用直接展开法推导过“广义二项式定理”,即有如下的泰勒展开:

%5Cbbox%5B%238ffa9a%5D%7B(1%2Bx)%5E%5Calpha%20%3D1%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20C_%7B%5Calpha%20%7D%5E%7Bk%7Dx%5Ek%2B%5Comicron%20(x%5En)%7D

C_%7B%5Calpha%20%7D%5Ek%3D%5Cfrac%7B%5Calpha%20(%5Calpha%20-1)%5Ccdots%20(%5Calpha%20-(k-1))%7D%7Bk!%7D%20%2C%5Calpha%20%5Cin%20%5Cmathbb%7BR%7D%2Ck%5Cin%20%5Cmathbb%7BN%7D%5E*广义组合数

应用(1):间接法展开ln(1+x)

取α=-1得:

%5Cbegin%7Balign%7D%0A%5Cfrac%7B1%7D%7B1%2Bx%7D%20%20%26%3D1%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20(-1)%5Ekx%5Ek%2B%5Comicron%20(x%5En)%5C%5C%0A%26%3D1-x%2Bx%5E2-%5Ccdots%2B(-1)%5Enx%5En%2B%5Comicron%20(x%5En)%5C%5C%0A%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ekx%5Ek%2B%5Comicron%20(x%5En)%0A%5Cend%7Balign%7D

由定理3得:

%5Cbegin%7Balign%7D%0A%5Cint_%7B0%7D%5E%7Bx%7D%20%5Cfrac%7B1%7D%7B1%2Bt%7D%5Cmathrm%7Bd%7Dt%20%26%3D%0A%5Cint_%7B0%7D%5E%7Bx%7D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ekt%5Ek%5Cmathrm%7Bd%7Dt%2B%5Comicron%20(x%5E%7Bn%2B1%7D)%20%5C%5C%0A%5CRightarrow%20%5Cln%20(1%2Bx)%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ek%5Cfrac%7Bx%5E%7Bk%2B1%7D%7D%7Bk%2B1%7D%20%2B%5Comicron%20(x%5E%7Bn%2B1%7D)%20%0A%5Cend%7Balign%7D

ps:对于有限个函数之和,求导和求和可以交换次序(这是使用了导数的加法法则)

由此可见,ln(1+x)的展开式也可以用间接法推得

应用(2):间接法展开arctanx和artanhx

由应用(1)我们已经得出:

%5Cbegin%7Balign%7D%0A%5Cfrac%7B1%7D%7B1%2Bx%7D%20%20%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ekx%5Ek%2B%5Comicron%20(x%5En)%0A%5Cend%7Balign%7D

由定理2,将上式的x换为x²得:

%5Cfrac%7B1%7D%7B1%2Bx%5E2%7D%20%20%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ekx%5E%7B2k%7D%2B%5Comicron%20(x%5E%7B2n%7D)

再由定理3得:

%5Cbegin%7Balign%7D%0A%5Cint_%7B0%7D%5E%7Bx%7D%20%5Cfrac%7B1%7D%7B1%2Bt%5E2%7D%5Cmathrm%7Bd%7Dt%20%20%26%20%3D%5Cint_%7B0%7D%5E%7Bx%7D%20%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ekt%5E%7B2k%7D%5Cmathrm%7Bd%7Dt%20%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%5C%5C%0A%5CRightarrow%20%5Carctan%20x%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20(-1)%5Ek%5Cfrac%7Bx%5E%7B2k%2B1%7D%7D%7B2k%2B1%7D%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%0A%5Cend%7Balign%7D

这样我们就推出了arctanx的泰勒展开式了!


类似的,如果将x换成-x²,同理可得artanhx的泰勒展开式:

%5Cfrac%7B1%7D%7B2%7D%20%5Cln%20(%5Cfrac%7B1%2Bx%7D%7B1-x%7D%20)%3D%5Cmathrm%7Bartanh%7D%20%20x%3D%5Csum_%7Bk%3D0%7D%5E%7Bn%7D%20%5Cfrac%7Bx%5E%7B2k%2B1%7D%7D%7B2k%2B1%7D%2B%5Comicron%20(x%5E%7B2n%2B1%7D)

应用(3):间接法展开arcsinx和arsinhx

广义二项式定理中取α=-1/2并将x换为x²得:

%5Cfrac%7B1%7D%7B%5Csqrt%7B1%2Bx%5E2%7D%20%7D%20%3D1%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20C_%7B-%5Cfrac%7B1%7D%7B2%7D%20%7D%5E%7Bk%7Dx%5E%7B2k%7D%2B%5Comicron%20(x%5E%7B2n%7D)

其中

%5Cbegin%7Balign%7D%0AC_%7B-%5Cfrac%7B1%7D%7B2%7D%20%7D%5Ek%26%3D%5Cfrac%7B(-%5Cfrac%7B1%7D%7B2%7D%20)(-%5Cfrac%7B1%7D%7B2%7D%20-1)(-%5Cfrac%7B1%7D%7B2%7D-2%20)%5Ccdots%20(-%5Cfrac%7B1%7D%7B2%7D-(k-1)%20)%7D%7Bk!%7D%5C%5C%0A%26%3D(-1)%5Ek%5Cfrac%7B1%5Ccdot%203%5Ccdot%205%5Ccdots%20(2k-1)%7D%7B2%5Ek%20k!%7D%5C%5C%0A%26%3D(-1)%5Ek%5Cfrac%7B1%5Ccdot%203%5Ccdot%205%5Ccdots%20(2k-1)%7D%7B2%5Ccdot%204%5Ccdot%206%5Ccdots%20(2k)%7D%5C%5C%0A%26%3D%20(-1)%5Ek%5Cfrac%7B(2k-1)!!%7D%7B(2k)!!%7D%20%0A%5Cend%7Balign%7D

ps:!!这个是双阶乘符号,表示隔一个整数跳着乘

对于奇数,定义:(2k-1)!!=(2k-1)*(2k-3)*...*5*3*1

对于偶数,定义:(2k)!!=(2k)*(2k-2)*...*6*4*2

代回上式得:

%5Cfrac%7B1%7D%7B%5Csqrt%7B1%2Bx%5E2%7D%20%7D%20%3D1%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20%20(-1)%5Ek%5Cfrac%7B(2k-1)!!%7D%7B(2k)!!%7D%20x%5E%7B2k%7D%2B%5Comicron%20(x%5E%7B2n%7D)

再由定理3得:

%5Cbegin%7Balign%7D%0A%5Cint_%7B0%7D%5E%7Bx%7D%20%5Cfrac%7B1%7D%7B%5Csqrt%7B1%2Bt%5E2%7D%20%7D%5Cmathrm%7Bd%7Dt%20%20%26%3D%5Cint_%7B0%7D%5E%7Bx%7D%5B1%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20%20(-1)%5Ek%5Cfrac%7B(2k-1)!!%7D%7B(2k)!!%7D%20t%5E%7B2k%7D%5D%5Cmathrm%7Bd%7Dt%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%5C%5C%0A%5CRightarrow%20%5Cmathrm%7Barsinh%7D%20x%26%3Dx%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20%20(-1)%5Ek%5Cfrac%7B(2k-1)!!%7D%7B(2k)!!(2k%2B1)%7D%20x%5E%7B2k%2B1%7D%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%0A%5Cend%7Balign%7D

这样我们就推出了arsinhx的泰勒展开式了!

ps:arsinhx是反双曲正弦函数,其对数函数的形式即:

%5Cmathrm%7Barsinh%7D%20x%3D%5Cln%20(x%2B%5Csqrt%7B1%2Bx%5E2%7D%20)


类似的,如果将x换成-x²,同理可得:

%5Carcsin%20x%3Dx%2B%5Csum_%7Bk%3D1%7D%5E%7Bn%7D%20%20%5Cfrac%7B(2k-1)!!%7D%7B(2k)!!(2k%2B1)%7D%20x%5E%7B2k%2B1%7D%2B%5Comicron%20(x%5E%7B2n%2B1%7D)

总结

本篇文章介绍了展开函数的3个重要的间接法,即前文定理1(函数乘积的展开法),定理2(复合函数的展开法),定理3(导函数/原函数的展开法)

我们将初等函数的泰勒展开选阶准则总结如下:

(1)对于函数加法和数乘展开准则(各自展开到n阶)

将g(x)=c₁f₁(x)+c₂f₂(x)+...+cₘfₘ(x)展开到n阶,需要将

f₁(x),f₂(x),...,fₘ(x)各自都展开到n阶

(2)对于函数乘法展开准则(个因子展开阶数=n-其余因子最低阶之和)

将g(x)=f₁(x)*f₂(x)*...*fₘ(x)展开到n阶,记f₁(x),f₂(x),...,fₘ(x)的(最低)阶数分别为:α₁,α₂,...,αₘ,则:

f₁(x)需展开到第(n-α₂-α₃-...-αₘ)阶;

f₂(x)需展开到第(n-α₁-α₃-...-αₘ)阶;

f₃(x)需展开到第(n-α₁-α₂-α₄...-αₘ)阶;

...

fₘ(x)需展开到第(n-α₁-α₂-...-αₘ₋₁)阶

(3)对于函数复合展开准则(先展外层,再逐个展开内层)

用□代表内层函数,则有:

%5Cbbox%5B%238ffa9a%5D%7B%5CBox%20%5Cto%20x_0%2Cf(%5CBox)%3D%5Csum_%7Bk%3D0%7D%5E%7Bm%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(%5CBox-x_0)%5Ek%20%20%2B%5Comicron%20%20((%5CBox-x_0)%5Em)%7D

记□-x₀的阶数为α,则(□-x₀)ᵐ的阶数为mα

要使得余项阶数≥n,即需mα≥n,则外层需要展开的阶数为:m=⌊ n/α ⌋

展开了外层后,再将每个%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(%5CBox-x_0)%5Ek%20按函数乘法法则展开准则逐个展开

(4)对于函数的导函数/原函数展开

f'(x)要展开到n阶,则需将f(x)展开到n+1阶再对Tₙ₊₁(x)“逐项求导”;

F(x)要展开到n阶,则需将f(x)展开到n-1阶再对Tₙ₋₁(x)“逐项积分


%5Cbegin%7Balign%7D%0Af(x)%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bm%2B1%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bm%2B1%7D)%20%5C%5C%0A%5CRightarrow%20f'(x)%26%3D%20%5Cfrac%7B%5Cmathrm%7Bd%7D%20%7D%7B%5Cmathrm%7Bd%7D%20x%7D%20%5B%5Csum_%7Bk%3D0%7D%5E%7Bm%2B1%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%5D%2B%20%2B%5Comicron%20%20((x-x_0)%5E%7Bm%7D)%20%0A%5Cend%7Balign%7D


%5Cbegin%7Balign%7D%0Af(x)%26%3D%5Csum_%7Bk%3D0%7D%5E%7Bm-1%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(x-x_0)%5Ek%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bm-1%7D)%20%5C%5C%0A%5CRightarrow%20%5Cint_%7Bx_0%7D%5E%7Bx%7D%20f(t)%5Cmathrm%7Bd%7Dt%20%26%3D%20%5Cint_%7Bx_0%7D%5E%7Bx%7D%20%5B%5Csum_%7Bk%3D0%7D%5E%7Bm-1%7D%5Cfrac%7Bf%5E%7B(k)%7D(x_0)%7D%7Bk!%7D(t-x_0)%5Ek%5D%5Cmathrm%7Bd%7Dt%20%20%2B%5Comicron%20%20((x-x_0)%5E%7Bm%7D)%20%0A%5Cend%7Balign%7D

这4个选阶的准则务必配加相应的练习进行巩固,才能熟练地选择到合适的阶数。

并且在最后用间接展开法推出了如下几个新的泰勒公式:

%5Cbbox%5B%238ffa9a%5D%7B%5Cbegin%7Balign%7D%0A%5Carctan%20x%26%3Dx-%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Cfrac%7Bx%5E5%7D%7B5%7D-%5Ccdots%2B%20%5Cfrac%7B(-1)%5En%20x%5E%7B2n%2B1%7D%7D%7B2n%2B1%7D%20%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%5C%5C%0A%5Cmathrm%7Bartanh%7D%20%20x%26%3Dx%2B%5Cfrac%7Bx%5E3%7D%7B3%7D%2B%5Cfrac%7Bx%5E5%7D%7B5%7D%2B%5Ccdots%20%2B%5Cfrac%7Bx%5E%7B2n%2B1%7D%20%7D%7B2n%2B1%7D%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%5C%5C%0A%5Carcsin%20x%26%3Dx%2B%5Cfrac%7B1!!%7D%7B2!!%5Ccdot%203%7D%20x%5E%7B3%7D%2B%5Cfrac%7B3!!%7D%7B4!!%5Ccdot%205%7D%20x%5E%7B5%7D%2B%5Ccdots%2B%5Cfrac%7B(2n-1)!!%7D%7B(2n)!!(2n%2B1)%7D%20x%5E%7B2n%2B1%7D%20%20%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%5C%5C%0A%5Cmathrm%7Barsinh%7D%20%20x%26%3Dx-%5Cfrac%7B1!!%7D%7B2!!%5Ccdot%203%7D%20x%5E%7B3%7D%2B%5Cfrac%7B3!!%7D%7B4!!%5Ccdot%205%7D%20x%5E%7B5%7D-%5Ccdots%2B%5Cfrac%7B(-1)%5En(2n-1)!!%7D%7B(2n)!!(2n%2B1)%7D%20x%5E%7B2n%2B1%7D%20%20%2B%5Comicron%20(x%5E%7B2n%2B1%7D)%0A%5Cend%7Balign%7D%7D


ps:三角函数和双曲函数是一对孪生兄弟,因此上面写成反双曲函数形式上更整齐些。这两者的对数形式如下:

%5Cmathrm%7Bartanh%7D%20%20x%3D%5Cfrac%7B1%7D%7B2%7D%20%5Cln%20(%5Cfrac%7B1%2Bx%7D%7B1-x%7D%20)%2C%0A%5Cmathrm%7Barsinh%7D%20%20x%3D%5Cln%20(x%2B%5Csqrt%7B1%2Bx%5E2%7D%20)

%5Cfrac%7B(2n-1)!!%7D%7B(2n)!!(2n%2B1)%7Dx%5E%7B2n%2B1%7D%20中分母2n+1对应于该项系数,然后把系数减1再双阶乘(2n)!!乘到分母,最后把系数减2再双阶乘(2n-1)!!乘到分子。也就是这里的2n+1,2n,2n-1刚好成等差数列,记下来其实也还不算太麻烦,并且注意一下哪几个有正负交错就好了~

至此,有关(带佩亚诺余项)的泰勒公式的知识点就分享完啦!在下一篇文章中,笔者会在题库中选择一些例题带读者们进行练习,巩固前篇文章和本篇文章提及的知识点~


Copyright © 2025 aigcdaily.cn  北京智识时代科技有限公司  版权所有  京ICP备2023006237号-1