微分学＜2＞——导数的定义与求导法则

导数

导数的定义

定义2.1 导数

若$y=f\left(x\right)$可微,则当$\Delta x\to 0$时,有${\lim }_{\Delta x\to 0}\left(\frac{\Delta y}{\Delta x}\right)={\lim }_{\Delta x\to 0}\left(\frac{f\left(x+△x\right)-f\left(x\right)}{\Delta x}\right)$存在且有限,则称函数$y=f\left(x\right)$可导,称${\lim }_{\Delta x\to 0}\left(\frac{\Delta y}{\Delta x}\right)$为$f\left(x\right)$的导函数$f^{\prime }\left(x\right)=\frac{dy}{dx}=\frac{df}{dx}$,函数$f\left(x\right)$在$x_0$处的导函数值称为$f^{\prime }\left(x_0\right)={\frac{dy}{dx}|}_{x=x_0}={\frac{df}{dx}|}_{x=x_0}$。
若函数$f\left(x\right)$在开区间$\left(a,b\right)$上每一点处处可导,则称函数$f\left(x\right)$在开区间$\left(a,b\right)$上可导。

导数的几何意义

切线斜率

一元函数$y=f\left(x\right)$在点$x_0$处的导数值,其几何意义为一元函数$y=f\left(x\right)$的图像上,过点$x_0$的切线斜率,同样也可以结合实际理解成变化率(物理中的变化率,经济中的边际效益等实际问题都需要考虑变化率),点$x_0$处的切线点斜式方程为:
$y-y_0=f^{\prime }\left(x_0\right)\left(x-x_0\right)$。
而${\lim }_{\Delta x\to 0}\left(\frac{\Delta y}{\Delta x}\right)={\lim }_{\Delta x\to 0}\left(\frac{f\left(x_0+△x\right)-f\left(x_0\right)}{\Delta x}\right)$可视为连接两点$\left(x_0,y_0\right)$,$\left(x_0+\Delta x,y_0+\Delta y\right)$的割线,当$\Delta x\to 0$时,割线向切线无限趋近,割线斜率的极限值为$y=f\left(x\right)$过点$x_0$的切线斜率。

左右导数

定义2.2 单侧导数

${\lim }_{\Delta x\to 0^{-}}\left(\frac{\Delta y}{\Delta x}\right)={\lim }_{\Delta x\to 0^{-}}\left(\frac{f\left(x_0+△x\right)-f\left(x_0\right)}{\Delta x}\right)$称为$f\left(x\right)$的左导数,记为$f_{-}\left(x_0\right)$。
${\lim }_{\Delta x\to 0^{+}}\left(\frac{\Delta y}{\Delta x}\right)={\lim }_{\Delta x\to 0^{+}}\left(\frac{f\left(x_0+△x\right)-f\left(x_0\right)}{\Delta x}\right)$称为$f\left(x\right)$的右导数,记为$f_{+}\left(x_0\right)$。
若函数$f\left(x\right)$在开区间$\left(a,b\right)$上可导,在左端点$a$处右导数$f_{+}\left(a\right)$存在,在右端点$b$处左导数$f_{-}\left(b\right)$存在,则称函数$f\left(x\right)$在闭区间$\left[a,b\right]$上可导。
导数存在当且仅当左右导数存在,有限且相等。
若点$x_0$处的左右导数存在,有限,但不相等,此时函数图像中$x_0$处出现尖锐的拐点,导函数图像中$x_0$处出现跳跃间断点,反映在物理世界中,就是变化率瞬间出现突变,就好像电子从低能级轨道瞬时跃迁到高能级轨道并放出能量(光子)一样,中间并不需要在两个能级中的空间中移动。
下面介绍一种处处连续而处处不可导的Weierstrass函数,Weierstrass函数呈现一种"分形"的特点,在当前尺度下,Weierstrass函数图像中会出现一些较大的尖角,较小的锯齿,和相对比较"平滑"的曲线,而将看起来"平滑"的曲线局部放大(或者提高图像精度),函数曲线会自动生成相似的尖角填补原先曲线的细节,以此类推,任何大尺度下的平滑处都能在小尺度上复现自身整体的崎岖,这种整体与部分的自相似性使得Weierstrass函数的每一处均不可导。

导数的四则运算

定理2.1 导数的四则运算法则

(1)${\left(\alpha f\left(x\right)\pm \beta g\left(x\right)\right)}^{\prime }=\alpha f^{\prime }\left(x\right)\pm \beta g^{\prime }\left(x\right)$
(2)${\left(f\left(x\right)g\left(x\right)\right)}^{\prime }=f^{\prime }\left(x\right)g\left(x\right)+f\left(x\right)g^{\prime }\left(x\right)$
(3)${\left(\frac{f\left(x\right)}{g\left(x\right)}\right)}^{\prime }(g\left(x\right)\ne 0)=\frac{f^{\prime }\left(x\right)g\left(x\right)-f\left(x\right)g^{\prime }\left(x\right)}{g^2\left(x\right)}$

(1)
以加法为例:
$$\begin{array}{lc}& {\left(\alpha f\left(x\right)+\beta g\left(x\right)\right)}^{\prime }\\ =& {\lim }_{\Delta x\to 0}\left(\frac{\left(\alpha f\left(x+\Delta x\right)+\beta g\left(x+\Delta x\right)\right)-\left(\alpha f\left(x\right)+\beta g\left(x\right)\right)}{\Delta x}\right)\\ =& {\lim }_{\Delta x\to 0}\left(\alpha \left(\frac{f\left(x+\Delta x\right)-f\left(x\right)}{\Delta x}\right)+\beta \left(\frac{g\left(x+\Delta x\right)-g\left(x\right)}{\Delta x}\right)\right)\\ =& \alpha \left({\lim }_{\Delta x\to 0}\left(\frac{f\left(x+\Delta x\right)-f\left(x\right)}{\Delta x}\right)\right)+\beta \left({\lim }_{\Delta x\to 0}\left(\frac{g\left(x+\Delta x\right)-g\left(x\right)}{\Delta x}\right)\right)\\ =& \alpha f^{\prime }\left(x\right)+\beta g^{\prime }\left(x\right)\end{array}$$
(2)
$$\begin{array}{lc}& {\left(f\left(x\right)\cdot g\left(x\right)\right)}^{\prime }\\ =& {\lim }_{\Delta x\to 0}\left[\frac{f\left(x+\Delta x\right)g\left(x+\Delta x\right)-f\left(x\right)g\left(x\right)}{\Delta x}\right]\\ =& {\lim }_{\Delta x\to 0}\left[\frac{f\left(x+\Delta x\right)g\left(x+\Delta x\right)-f\left(x\right)g\left(x+\Delta x\right)+f\left(x\right)g\left(x+\Delta x\right)-f\left(x\right)g\left(x\right)}{\Delta x}\right]\\ =& {\lim }_{\Delta x\to 0}g\left(x+\Delta x\right)\cdot {\lim }_{\Delta x\to 0}\frac{f\left(x+\Delta x\right)-f\left(x\right)}{\Delta x}+f\left(x\right)\cdot {\lim }_{\Delta x\to 0}\frac{g\left(x+\Delta x\right)-g\left(x\right)}{\Delta x}\\ =& f^{\prime }\left(x\right)g\left(x\right)+f\left(x\right)g^{\prime }\left(x\right)\end{array}$$
(3)
先证明${\left(\frac{1}{g\left(x\right)}\right)}^{\prime }=-\frac{g^{\prime }\left(x\right)}{g^2\left(x\right)}$:
$$\begin{array}{lc}& {\left(\frac{1}{g\left(x\right)}\right)}^{\prime }\\ =& {\lim }_{\Delta x\to 0}\frac{\frac{1}{g\left(x+\Delta x\right)}-\frac{1}{g\left(x\right)}}{\Delta x}\\ =& {\lim }_{\Delta x\to 0}\frac{g\left(x\right)-g\left(x+\Delta x\right)}{g\left(x+\Delta x\right)g\left(x\right)\Delta x}\\ =& {\lim }_{\Delta x\to 0}\frac{1}{g\left(x\right)g\left(x+\Delta x\right)}\cdot {\lim }_{\Delta x\to 0}\frac{g\left(x\right)-g\left(x+\Delta x\right)}{\Delta x}\\ =& -\frac{g^{\prime }\left(x\right)}{g^2\left(x\right)}\end{array}$$
$$\begin{array}{lc}& {\left(\frac{f\left(x\right)}{g\left(x\right)}\right)}^{\prime }\\ =& {\left(f\left(x\right)\cdot {\left(g\left(x\right)\right)}^{-1}\right)}^{\prime }\\ =& f^{\prime }\left(x\right)\cdot {\left(g\left(x\right)\right)}^{-1}+f\left(x\right)\cdot {\left[{\left(g\left(x\right)\right)}^{-1}\right]}^{\prime }\\ =& \frac{f^{\prime }\left(x\right)g\left(x\right)-f\left(x\right)g^{\prime }\left(x\right)}{g^2\left(x\right)}\end{array}$$

反函数的导数

定理2.2 反函数求导法则

若函数$f\left(x\right)\ne 0$在开区间$\left(a,b\right)$上连续,严格单调增加(减少)且可导,设 α = min ⁡ { f ( a + ) , f ( b − ) } \alpha =\min\left \{ f\left ( a^{+} \right ),f\left ( b^{-} \right ) \right \} α=min{f(a+),f(b−)}, β = max ⁡ { f ( a + ) , f ( b − ) } \beta =\max \left \{ f\left ( a^{+} \right ),f\left ( b^{-} \right ) \right \} β=max{f(a+),f(b−)}则反函数$x=f^{-1}\left(y\right)$在开区间$\left(\alpha ,\beta \right)$上可导,且$x=f^{-1}\left(y\right)$的导数为$\frac{d{f}^{-1}}{dy}=\frac{1}{f^{\prime }\left(x\right)}=\frac{1}{\frac{df}{dx}}$。

由反函数存在定理,$f\left(x\right)$的反函数$x=f^{-1}\left(y\right)$存在,且保持函数$f\left(x\right)$的严格单调性,$\Delta y\ne 0\iff \Delta x\ne 0$,
根据反函数连续性定理,$f\left(x\right)$的反函数$x=f^{-1}\left(y\right)$在$\left[\alpha ,\beta \right]$上连续,由连续函数定义,$\Delta y\to 0\iff \Delta x\to 0$,
根据导数定义,$\frac{d{f}^{-1}}{dx}={\lim }_{\Delta y\to 0}\frac{f^{-1}\left(y+\Delta y\right)-f^{-1}\left(y\right)}{\Delta y}$,
可得出以下关系:
$$\{\begin{array}{c}f\left(x+\Delta x\right)=y+\Delta y\\ f^{-1}\left(y+\Delta y\right)=x+\Delta x\end{array}$$
$$\begin{array}{lc}& \frac{d{f}^{-1}}{dx}\\ =& {\lim }_{\Delta x\to 0}\left(\frac{x+\Delta x-x}{f\left(x+\Delta x\right)-f\left(x\right)}\right)\\ =& {\lim }_{\Delta x\to 0}\left(\frac{\Delta x}{f\left(x+\Delta x\right)-f\left(x\right)}\right)\\ =& {\lim }_{\Delta x\to 0}\left(\frac{1}{\frac{f\left(x+△x\right)-f\left(x\right)}{\Delta x}}\right)\\ =& \frac{1}{\frac{df}{dx}}\end{array}$$
于是上述结论成立。

复合函数的导数

定理2.3 复合函数求导法则

$u=g\left(x\right)$在$x=x_0$可导,$g\left(x_0\right)=u_0$,同时$y=f\left(u\right)在u=u_0$处可导,则$f\left(g\left(x\right)\right)在x=x_0$可导,且${\left(f\left(g\left(x_0\right)\right)\right)}^{\prime }=\left(\frac{dy}{du}\cdot \frac{du}{dx}\right){\mid }_{x=x_0}$。

由可导的定义,$y=f\left(u\right)在u=u_0$处可微,将其表示为$\Delta y=f\left(u_0+\Delta u\right)-f\left(u_0\right)=f^{\prime }\left(u_0\right)\Delta u+o\left(\Delta u\right)$,
构造函数
$$\alpha =\{\begin{array}{cc}\frac{o\left(\Delta u\right)}{\Delta u}& \Delta u\ne 0\\ 0& \Delta u=0\end{array}$$
满足${\lim }_{\Delta u\to 0}\alpha =0$。
由连续函数定义,$\Delta u\to 0\iff \Delta x\to 0$,使得即使当$\Delta u=0$时,${\lim }_{\Delta x\to 0}\frac{\Delta y}{\Delta x}={\lim }_{\Delta x\to 0}\left(f^{\prime }\left(u_0\right)\cdot \frac{\Delta u}{\Delta x}+\frac{\alpha \Delta u}{\Delta x}\right)=f^{\prime }\left(u_0\right)\cdot {\lim }_{\Delta x\to 0}\left(\frac{\Delta u}{\Delta x}\right)=\left(\frac{dy}{du}\cdot \frac{du}{dx}\right){\mid }_{x=x_0}$。