目录
- n 次幂函数导数公式的推导
- 导数和的运算法则的证明
- 正弦、余弦函数导数公式的推导
- 代数证明
- 两个重要极限(引理)及证明
- 具体推导
- 几何直观
- 导数积的运算法则的证明
- 导数商的法则的证明
- 链式法则的证明
- 有理幂函数求导法则的证明
- 反函数求导法则的证明
- 反正切函数导数公式的推导
- 指数函数导数公式的推导
- 引入
- 证明
- 第一种方法:代入消元
- 第二种方法:对数微分
- 幂指函数导数公式的推导
- 幂法则(The Power Rule)的证明
- 第一种方法:e 的代换
- 第二种方法:对数微分
- e 的极限本质
- 双曲函数导数公式的推导
- 重要性质的证明
非专业、适用于计算机科学的微积分简要提纲,用于汇总梳理知识。如有错误恳请批评指正!
n 次幂函数导数公式的推导
d
d
x
x
n
=
lim
Δ
x
→
0
Δ
y
Δ
x
=
lim
Δ
x
→
0
(
x
+
Δ
x
)
n
−
x
n
Δ
x
\begin{aligned} \frac{d}{dx} x^n & = \lim_{\Delta x\to 0} \frac{\Delta y}{\Delta x} \\\\ & = \lim_{\Delta x\to 0} \frac{(x+\Delta x)^n - x^n}{\Delta x} \end{aligned}
dxdxn=Δx→0limΔxΔy=Δx→0limΔx(x+Δx)n−xn
对于微小量
Δ
x
\Delta x
Δx,根据牛顿二项式定理
( x + y ) n = ∑ k = 0 n ( n k ) x n − k y k (x+y)^n = \sum_{k=0}^{n} \binom{n}{k} x^{n-k} y^k (x+y)n=k=0∑n(kn)xn−kyk
我们有(略去无穷小量余项)
(
x
+
Δ
x
)
n
=
x
n
+
n
(
Δ
x
)
x
n
−
1
+
o
(
Δ
x
)
2
(x+\Delta x)^n = x^n + n(\Delta x) x^{n-1} + o(\Delta x)^2
(x+Δx)n=xn+n(Δx)xn−1+o(Δx)2
则有
Δ
y
Δ
x
=
[
x
n
+
n
(
Δ
x
)
x
n
−
1
+
o
(
Δ
x
)
2
]
−
x
n
Δ
x
=
n
x
n
−
1
+
o
(
Δ
x
)
\begin{aligned} \frac{\Delta y}{\Delta x} & = \frac{[x^n + n(\Delta x) x^{n-1} + o(\Delta x)^2] -x^n}{\Delta x} \\\\ &= nx^{n-1} + o(\Delta x) \end{aligned}
ΔxΔy=Δx[xn+n(Δx)xn−1+o(Δx)2]−xn=nxn−1+o(Δx)
当
Δ
x
→
0
\Delta x \to 0
Δx→0 时,我们有
lim
Δ
x
→
0
Δ
y
Δ
x
=
n
x
n
−
1
\lim_{\Delta x\to 0} \frac{\Delta y}{\Delta x} = nx^{n-1}
Δx→0limΔxΔy=nxn−1
即
d
d
x
x
n
=
n
x
n
−
1
\boxed{\frac{d}{dx} x^n = nx^{n-1}}
dxdxn=nxn−1
导数和的运算法则的证明
( u + v ) ′ ( x ) = u ′ ( x ) + v ′ ( x ) \boxed{(u+v)'(x) = u'(x) + v'(x)} (u+v)′(x)=u′(x)+v′(x)
( u + v ) ′ ( x ) = lim Δ x → 0 ( u + v ) ( x + Δ x ) − ( u + v ) ( x ) Δ x = lim Δ x → 0 u ( x + Δ x ) − u ( x ) + v ( x + Δ x ) − v ( x ) Δ x = lim Δ x → 0 u ( x + Δ x ) − u ( x ) Δ x + lim Δ x → 0 v ( x + Δ x ) − v ( x ) Δ x = u ′ ( x ) + v ′ ( x ) . \begin{aligned} (u+v)'(x) & = \lim_{\Delta x \to 0} \frac{(u+v)(x+\Delta x)-(u+v)(x)}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{u(x+\Delta x)-u(x)+v(x+\Delta x)-v(x)}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{u(x+\Delta x)-u(x)}{\Delta x} + \lim_{\Delta x \to 0} \frac{v(x+\Delta x)-v(x)}{\Delta x} \\\\ &= u'(x)+v'(x). \end{aligned} (u+v)′(x)=Δx→0limΔx(u+v)(x+Δx)−(u+v)(x)=Δx→0limΔxu(x+Δx)−u(x)+v(x+Δx)−v(x)=Δx→0limΔxu(x+Δx)−u(x)+Δx→0limΔxv(x+Δx)−v(x)=u′(x)+v′(x).
正弦、余弦函数导数公式的推导
代数证明
两个重要极限(引理)及证明
lim
Δ
x
→
0
c
o
s
Δ
x
−
1
Δ
x
=
0
\boxed{\lim_{\Delta x \to 0} \frac{cos\Delta x - 1}{\Delta x} = 0 }
Δx→0limΔxcosΔx−1=0
设
Δ
x
=
θ
\Delta x = \theta
Δx=θ,我们考虑一个半径
r
=
1
\ r = 1
r=1 的单位圆,以弧度为度量单位,
θ
\theta
θ 扫过的弧线长度
l
=
θ
r
=
θ
\ l = \theta r = \theta
l=θr=θ。过
θ
\theta
θ 角的终止角向其起始角作垂线,则垂点到圆心的水平距离为
c
o
s
θ
\ cos \ \theta
cos θ,垂点到弧的一端的水平距离为
1
−
c
o
s
θ
\ 1-cos\ \theta
1−cos θ。
当
θ
→
0
\theta \to 0
θ→0 时,
1
−
c
o
s
θ
\ 1-cos\ \theta
1−cos θ 递减比
θ
\theta
θ 更快,因此比值
1
−
c
o
s
θ
θ
→
0
\frac{1-cos\ \theta}{\theta} \to 0
θ1−cos θ→0,则可以证明
lim
Δ
x
→
0
c
o
s
Δ
x
−
1
Δ
x
=
0
\lim_{\Delta x \to 0} \frac{cos\Delta x - 1}{\Delta x} = 0
Δx→0limΔxcosΔx−1=0 成立。
lim
Δ
x
→
0
s
i
n
Δ
x
Δ
x
=
1
\boxed{\lim_{\Delta x \to 0} \frac{sin\Delta x}{\Delta x} = 1}
Δx→0limΔxsinΔx=1
同理,如图
当
θ
→
0
\theta \to 0
θ→0 时,
s
i
n
θ
≈
θ
\ sin\ \theta \approx \theta
sin θ≈θ 逐步趋向于精确,比值
s
i
n
θ
θ
→
1
\frac{sin\ \theta}{\theta} \to 1
θsin θ→1,因此极限
lim
Δ
x
→
0
s
i
n
Δ
x
Δ
x
=
1
\lim_{\Delta x \to 0} \frac{sin\Delta x}{\Delta x} = 1
Δx→0limΔxsinΔx=1 成立。
具体推导
d d x s i n x = lim Δ x → 0 s i n ( x + Δ x ) − s i n x Δ x = lim Δ x → 0 s i n x c o s Δ x + c o s x s i n Δ x − s i n x Δ x = lim Δ x → 0 [ s i n x c o s Δ x − s i n x Δ x + c o s x s i n Δ x Δ x ] = lim Δ x → 0 [ s i n x ( c o s Δ x − 1 Δ x ) + c o s x ( s i n Δ x Δ x ) ] \begin{aligned} \frac{d}{dx} sin\ x &= \lim_{\Delta x \to 0} \frac{sin(x+\Delta x)-sin\ x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{sin\ x\ cos\Delta x + cos\ x \ sin\Delta x - sin \ x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \left[\frac{sin\ x\ cos\Delta x - sin \ x}{\Delta x} + \frac{cos\ x \ sin\Delta x }{\Delta x}\right] \\\\ &= \lim_{\Delta x \to 0} \left[sin\ x\left(\frac{cos\Delta x - 1}{\Delta x}\right) + cos\ x\left( \frac{ \ sin\Delta x }{\Delta x}\right)\right] \\\\ \end{aligned} dxdsin x=Δx→0limΔxsin(x+Δx)−sin x=Δx→0limΔxsin x cosΔx+cos x sinΔx−sin x=Δx→0lim[Δxsin x cosΔx−sin x+Δxcos x sinΔx]=Δx→0lim[sin x(ΔxcosΔx−1)+cos x(Δx sinΔx)]
利用两个重要极限,可得
d
d
x
s
i
n
x
=
c
o
s
x
\begin{aligned} \boxed{\frac{d}{dx} sin\ x = cos\ x} \end{aligned}
dxdsin x=cos x
同样地
d
d
x
c
o
s
x
=
lim
Δ
x
→
0
c
o
s
(
x
+
Δ
x
)
−
c
o
s
x
Δ
x
=
lim
Δ
x
→
0
c
o
s
x
c
o
s
Δ
x
−
s
i
n
x
s
i
n
Δ
x
−
c
o
s
x
Δ
x
=
lim
Δ
x
→
0
[
c
o
s
x
c
o
s
Δ
x
−
c
o
s
x
Δ
x
−
s
i
n
x
s
i
n
Δ
x
Δ
x
]
=
lim
Δ
x
→
0
[
c
o
s
x
(
c
o
s
Δ
x
−
1
Δ
x
)
−
s
i
n
x
(
s
i
n
Δ
x
Δ
x
)
]
\begin{aligned} \frac{d}{dx} cos\ x &= \lim_{\Delta x \to 0} \frac{cos(x+\Delta x)-cos\ x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{cos\ x\ cos\Delta x - sin\ x \ sin\Delta x - cos \ x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \left[\frac{cos\ x\ cos\Delta x - cos \ x}{\Delta x} - \frac{sin\ x \ sin\Delta x }{\Delta x}\right] \\\\ &= \lim_{\Delta x \to 0} \left[cos\ x\left(\frac{cos\Delta x - 1}{\Delta x}\right) - sin\ x\left( \frac{ \ sin\Delta x }{\Delta x}\right)\right] \\\\ \end{aligned}
dxdcos x=Δx→0limΔxcos(x+Δx)−cos x=Δx→0limΔxcos x cosΔx−sin x sinΔx−cos x=Δx→0lim[Δxcos x cosΔx−cos x−Δxsin x sinΔx]=Δx→0lim[cos x(ΔxcosΔx−1)−sin x(Δx sinΔx)]
利用两个重要极限,可得
d
d
x
c
o
s
x
=
−
s
i
n
x
\begin{aligned} \boxed{\frac{d}{dx} cos\ x = -sin\ x} \end{aligned}
dxdcos x=−sin x
几何直观
我们在单位圆上给予
θ
\theta
θ 微小增量
Δ
θ
\Delta \theta
Δθ,则 P 的纵坐标为
s
i
n
θ
\ sin\ \theta
sin θ,Q 的纵坐标为
s
i
n
(
θ
+
Δ
θ
)
\ sin(\theta + \Delta \theta)
sin(θ+Δθ)。
如图,纵坐标增量
Δ
y
=
∣
P
R
∣
\Delta y = |PR|
Δy=∣PR∣,线段 PQ 则是对弧线
P
Q
⌢
\overset{\LARGE{\frown}}{PQ}
PQ⌢的近似,我们可以很容易得到
∣
P
Q
∣
≈
Δ
θ
\ |PQ| \approx \Delta\theta
∣PQ∣≈Δθ。考虑到
Δ
θ
\Delta \theta
Δθ 为微小量,线段 PQ 几乎是圆的一条切线,以至于
∠
O
P
Q
\angle OPQ
∠OPQ 近似为直角,由几何关系,
∠
R
P
Q
≈
θ
\angle RPQ \approx \theta
∠RPQ≈θ,则
c
o
s
θ
≈
∣
P
R
∣
Δ
θ
=
s
i
n
(
θ
+
Δ
θ
)
−
s
i
n
θ
Δ
θ
.
\ cos\ \theta \approx \frac{|PR|}{\Delta\theta}=\frac{sin(\theta+\Delta\theta)-sin\ \theta}{\Delta\theta}.
cos θ≈Δθ∣PR∣=Δθsin(θ+Δθ)−sin θ.
那么,当
Δ
θ
→
0
\Delta\theta \to 0
Δθ→0 时,这一近似结果就越准确,直到
lim
Δ
θ
→
0
s
i
n
(
θ
+
Δ
θ
)
−
s
i
n
θ
Δ
θ
=
c
o
s
θ
\lim_{\Delta\theta \to 0} \frac{sin(\theta+\Delta\theta)-sin\ \theta}{\Delta\theta} = cos\ \theta
Δθ→0limΔθsin(θ+Δθ)−sin θ=cos θ
成立。
y = c o s θ y = cos\ \theta y=cos θ 同理易证,但是需要注意余弦函数的单调性(单调递减,增量为负)。
导数积的运算法则的证明
( u v ) ′ = u ′ v + u v ′ \boxed{(uv)'=u'v+uv'} (uv)′=u′v+uv′
( u v ) ′ = lim Δ x → 0 u ( x + Δ x ) v ( x + Δ x ) − u ( x ) v ( x ) Δ x (uv)' = \lim_{\Delta x \to 0} \frac{u(x+\Delta x)v(x+\Delta x)-u(x)v(x)}{\Delta x} (uv)′=Δx→0limΔxu(x+Δx)v(x+Δx)−u(x)v(x)
我们希望构造出
[
u
(
x
+
Δ
x
)
−
u
(
x
)
]
v
(
x
)
\left[u(x+\Delta x) - u(x)\right]v(x)
[u(x+Δx)−u(x)]v(x) 这一项以期凑出
u
′
(
x
)
u'(x)
u′(x),于是我们注意到
u
(
x
+
Δ
x
)
v
(
x
)
−
u
(
x
+
Δ
x
)
v
(
x
)
=
0
u(x+\Delta x)v(x)-u(x+\Delta x)v(x)=0
u(x+Δx)v(x)−u(x+Δx)v(x)=0
代入得
(
u
v
)
′
=
lim
Δ
x
→
0
u
(
x
+
Δ
x
)
v
(
x
)
−
u
(
x
)
v
(
x
)
+
u
(
x
+
Δ
x
)
v
(
x
+
Δ
x
)
−
u
(
x
+
Δ
x
)
v
(
x
)
Δ
x
=
lim
Δ
x
→
0
{
[
u
(
x
+
Δ
x
)
−
u
(
x
)
Δ
x
]
v
(
x
)
+
u
(
x
+
Δ
x
)
[
v
(
x
+
Δ
x
)
−
v
(
x
)
Δ
x
]
}
=
u
′
(
x
)
v
(
x
)
+
lim
Δ
x
→
0
{
u
(
x
+
Δ
x
)
[
v
(
x
+
Δ
x
)
−
v
(
x
)
Δ
x
]
}
=
u
′
(
x
)
v
(
x
)
+
u
(
x
)
v
′
(
x
)
.
\begin{aligned} (uv)'&= \lim_{\Delta x \to 0} \frac{u(x+\Delta x)v(x)-u(x)v(x)+u(x+\Delta x)v(x+\Delta x)-u(x+\Delta x)v(x)}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \left\{ \left[\frac{u(x+\Delta x)-u(x)}{\Delta x}\right]v(x)+u(x+\Delta x) \left[\frac{v(x+\Delta x)-v(x)}{\Delta x}\right] \right\} \\\\ &= u'(x)v(x)+\lim_{\Delta x \to 0} \left\{u(x+\Delta x) \left[\frac{v(x+\Delta x)-v(x)}{\Delta x}\right] \right\} \\\\ &= u'(x)v(x)+u(x)v'(x). \end{aligned}
(uv)′=Δx→0limΔxu(x+Δx)v(x)−u(x)v(x)+u(x+Δx)v(x+Δx)−u(x+Δx)v(x)=Δx→0lim{[Δxu(x+Δx)−u(x)]v(x)+u(x+Δx)[Δxv(x+Δx)−v(x)]}=u′(x)v(x)+Δx→0lim{u(x+Δx)[Δxv(x+Δx)−v(x)]}=u′(x)v(x)+u(x)v′(x).
导数商的法则的证明
( u v ) ′ = u ′ v − u v ′ v 2 \boxed{\left(\frac{u}{v}\right)'=\frac{u'v-uv'}{v^2}} (vu)′=v2u′v−uv′
( u v ) ′ = lim Δ x → 0 u ( x + Δ x ) v ( x + Δ x ) − u ( x ) v ( x ) Δ x \begin{aligned} \left(\frac{u}{v}\right)' = \lim_{\Delta x \to 0} \frac{\frac{u(x+\Delta x)}{v(x+\Delta x)}-\frac{u(x)}{v(x)}}{\Delta x} \end{aligned} (vu)′=Δx→0limΔxv(x+Δx)u(x+Δx)−v(x)u(x)
设
Δ
u
=
u
(
x
+
Δ
x
)
−
u
(
x
)
\Delta u = u(x+\Delta x)-u(x)
Δu=u(x+Δx)−u(x) ,
Δ
v
=
v
(
x
+
Δ
x
)
−
v
(
x
)
\Delta v = v(x+\Delta x)-v(x)
Δv=v(x+Δx)−v(x) ,则有
u
(
x
+
Δ
x
)
v
(
x
+
Δ
x
)
−
u
(
x
)
v
(
x
)
=
u
+
Δ
u
v
+
Δ
v
−
u
v
=
(
u
+
Δ
u
)
v
−
u
(
v
+
Δ
v
)
(
v
+
Δ
v
)
v
=
u
v
+
(
Δ
u
)
v
−
u
v
+
u
(
Δ
v
)
(
v
+
Δ
v
)
v
=
(
Δ
u
)
v
+
u
(
Δ
v
)
(
v
+
Δ
v
)
v
\begin{aligned} \frac{u(x+\Delta x)}{v(x+\Delta x)}-\frac{u(x)}{v(x)} &= \frac{u+\Delta u}{v+\Delta v} - \frac{u}{v} \\\\ &= \frac{(u+\Delta u)v-u(v+\Delta v)}{(v+\Delta v)v} \\\\ &= \frac{uv+(\Delta u)v-uv+u(\Delta v)}{(v+\Delta v)v} \\\\ &= \frac{(\Delta u)v+u(\Delta v)}{(v+\Delta v)v} \end{aligned}
v(x+Δx)u(x+Δx)−v(x)u(x)=v+Δvu+Δu−vu=(v+Δv)v(u+Δu)v−u(v+Δv)=(v+Δv)vuv+(Δu)v−uv+u(Δv)=(v+Δv)v(Δu)v+u(Δv)
我们有
(
u
v
)
′
=
lim
Δ
x
→
0
(
Δ
u
)
v
+
u
(
Δ
v
)
(
v
+
Δ
v
)
u
Δ
x
=
lim
Δ
x
→
0
1
Δ
x
(
Δ
u
)
v
+
u
(
Δ
v
)
(
v
+
Δ
v
)
u
=
lim
Δ
x
→
0
(
Δ
u
Δ
x
)
v
−
u
(
Δ
v
Δ
x
)
(
v
+
Δ
v
)
v
=
(
d
u
d
x
)
v
−
u
(
d
v
d
x
)
v
2
=
u
′
v
−
u
v
′
v
2
.
\begin{aligned} \left(\frac{u}{v}\right)' &= \lim_{\Delta x \to 0} \frac{\frac{(\Delta u)v+u(\Delta v)}{(v+\Delta v)u}}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{1}{\Delta x} \frac{(\Delta u)v+u(\Delta v)}{(v+\Delta v)u} \\\\ &= \lim_{\Delta x \to 0} \frac{\left(\frac{\Delta u}{\Delta x}\right) v-u\left(\frac{\Delta v}{\Delta x}\right)}{(v+\Delta v)v} \\\\ &= \frac{ \left( \frac{du}{dx}\right)v-u \left( \frac{dv}{dx}\right)}{v^2} \\\\ &= \frac{u'v-uv'}{v^2}. \end{aligned}
(vu)′=Δx→0limΔx(v+Δv)u(Δu)v+u(Δv)=Δx→0limΔx1(v+Δv)u(Δu)v+u(Δv)=Δx→0lim(v+Δv)v(ΔxΔu)v−u(ΔxΔv)=v2(dxdu)v−u(dxdv)=v2u′v−uv′.
链式法则的证明
d y d t = d y d x d x d t \boxed{\frac{dy}{dt}= \frac{dy}{dx} \frac{dx}{dt}} dtdy=dxdydtdx
d y = d y d x d x = d y d x d x d t d t . \begin{aligned} dy &= \frac{dy}{dx} dx \\\\ &= \frac{dy}{dx} \frac{dx}{dt} dt. \end{aligned} dy=dxdydx=dxdydtdxdt.
有理幂函数求导法则的证明
若
a
∈
Q
a \in Q
a∈Q,则有
d
d
x
(
x
a
)
=
a
x
a
−
1
\boxed{\frac{d}{dx} (x^a)=ax^{a-1}}
dxd(xa)=axa−1
假设 a = m n a=\frac{m}{n} a=nm,其中 m ∈ N m \in N m∈N, n ∈ N n \in N n∈N,我们有 y = x m n y = x^{\frac{m}{n}} y=xnm,则
y n = ( x m n ) n y n = x m \begin{aligned} y^n &= \left(x^{\frac{m}{n}}\right)^n \\ y^n &= x^m \end{aligned} ynyn=(xnm)n=xm
等式两边同时求导
d
d
x
y
n
=
d
d
x
x
m
(
d
d
y
y
n
)
d
y
d
x
=
m
x
m
−
1
n
y
n
−
1
d
y
d
x
=
m
x
m
−
1
d
y
d
x
=
m
n
x
m
−
1
y
n
−
1
\begin{aligned} \frac{d}{dx} y^n &= \frac{d}{dx} x^m \\\\ \left(\frac{d}{dy} y^n\right) \frac{dy}{dx} &= mx^{m-1} \\\\ ny^{n-1}\ \frac{dy}{dx} &= mx^{m-1} \\\\ \frac{dy}{dx} &= \frac{m}{n} \frac{x^{m-1}}{y^{n-1}} \end{aligned}
dxdyn(dydyn)dxdynyn−1 dxdydxdy=dxdxm=mxm−1=mxm−1=nmyn−1xm−1
将
y
=
x
m
n
y = x^{\frac{m}{n}}
y=xnm 代入方程得
d
y
d
x
=
m
n
[
x
m
−
1
(
x
m
n
)
n
−
1
]
=
m
n
[
x
m
−
1
x
m
(
n
−
1
)
n
]
=
m
n
x
[
(
m
−
1
)
−
m
(
n
−
1
)
n
]
=
m
n
x
[
n
(
m
−
1
)
−
m
(
n
−
1
)
n
]
=
m
n
x
m
−
n
n
=
m
n
x
m
n
−
1
=
a
x
a
−
1
.
\begin{aligned} \frac{dy}{dx} &= \frac{m}{n} \left[ \frac{x^{m-1}}{(x^{\frac{m}{n}})^{n-1}}\right] \\\\ &= \frac{m}{n} \left[ \frac{x^{m-1}}{x^{\frac{m(n-1)}{n}}}\right] \\\\ &= \frac{m}{n} x^{\left[(m-1)-\frac{m(n-1)}{n}\right]} \\\\ &= \frac{m}{n} x^{\left[\frac{n(m-1)-m(n-1)}{n}\right]} \\\\ &= \frac{m}{n} x^{\frac{m-n}{n}} \\\\ &= \frac{m}{n} x^{\frac{m}{n}-1} \\\\ &= ax^{a-1}. \end{aligned}
dxdy=nm[(xnm)n−1xm−1]=nm[xnm(n−1)xm−1]=nmx[(m−1)−nm(n−1)]=nmx[nn(m−1)−m(n−1)]=nmxnm−n=nmxnm−1=axa−1.
反函数求导法则的证明
d x d y = 1 d y d x \boxed{\frac{dx}{dy}=\frac{1}{\frac{dy}{dx}}} dydx=dxdy1
我们有
y
=
f
(
x
)
y = f(x)
y=f(x),
f
−
1
(
y
)
=
x
f^{-1}(y)=x
f−1(y)=x,则
d
d
x
[
f
−
1
(
y
)
]
=
d
d
x
(
x
)
d
d
x
[
f
−
1
(
y
)
]
=
1
\begin{aligned} \frac{d}{dx}[f^{-1}(y)] &= \frac{d}{dx} (x) \\\\ \frac{d}{dx}[f^{-1}(y)] &= 1 \\\\ \end{aligned}
dxd[f−1(y)]dxd[f−1(y)]=dxd(x)=1
由链式法则
d
d
y
[
f
−
1
(
x
)
]
d
y
d
x
=
1
d
d
y
[
f
−
1
(
x
)
]
=
1
d
y
d
x
.
\begin{aligned} \frac{d}{dy}[f^{-1}(x)]\ \frac{dy}{dx} &= 1 \\\\ \frac{d}{dy}[f^{-1}(x)] &= \frac{1}{\frac{dy}{dx}}. \\\\ \end{aligned}
dyd[f−1(x)] dxdydyd[f−1(x)]=1=dxdy1.
反正切函数导数公式的推导
d d x a r c t a n ( x ) = 1 1 + x 2 \boxed{\frac{d}{dx}arctan(x) = \frac{1}{1+x^2}} dxdarctan(x)=1+x21
我们有
y
=
a
r
c
t
a
n
(
x
)
=
t
a
n
−
1
(
x
)
y = arctan(x)=tan^{-1}(x)
y=arctan(x)=tan−1(x) ,等式两边同时取正切以简化方程:
t
a
n
y
=
t
a
n
[
t
a
n
−
1
(
x
)
]
t
a
n
y
=
x
\begin{aligned} tan\ y &= tan[tan^{-1}(x)] \\ tan \ y &= x \end{aligned}
tan ytan y=tan[tan−1(x)]=x
如图,我们限制函数
t
a
n
(
x
)
tan(x)
tan(x) 的定义域为
(
−
π
2
,
π
2
)
\left(-\frac{\pi}{2}, \frac{\pi}{2}\right)
(−2π,2π) ,我们沿
y
=
x
y=x
y=x 将
t
a
n
(
x
)
tan(x)
tan(x) 翻转即可得到反函数
t
a
n
−
1
(
x
)
tan^{-1}(x)
tan−1(x) :
我们可以知道,
lim
x
→
∞
t
a
n
−
1
(
x
)
=
π
2
\lim_{x\to \infty} tan^{-1}(x) = \frac{\pi}{2}
limx→∞tan−1(x)=2π ,而我们显然可以推导出
d
d
y
t
a
n
y
=
d
d
y
s
i
n
y
c
o
s
y
=
c
o
s
2
y
−
s
i
n
2
y
c
o
s
2
y
=
1
c
o
s
2
y
=
s
e
c
2
y
\begin{aligned} \frac{d}{dy} tan\ y &= \frac{d}{dy} \frac{sin\ y}{cos\ y} \\\\ &= \frac{cos^2 y \ -sin^2y}{cos^2 y} \\\\ &= \frac{1}{cos^2 y} \\\\ &= sec^2 y \end{aligned}
dydtan y=dydcos ysin y=cos2ycos2y −sin2y=cos2y1=sec2y
我们对
t
a
n
y
=
x
tan\ y = x
tan y=x 两边同时求导
d
d
x
[
t
a
n
(
y
)
]
=
d
d
x
x
d
d
y
[
t
a
n
(
y
)
]
d
y
d
x
=
1
1
c
o
s
2
y
d
y
d
x
=
1
d
y
d
x
=
c
o
s
2
y
\begin{aligned} \frac{d}{dx}[tan(y)] &= \frac{d}{dx} x \\\\ \frac{d}{dy}[tan(y)]\ \frac{dy}{dx} &= 1 \\\\ \frac{1}{cos^2y}\ \frac{dy}{dx} &= 1 \\\\ \frac{dy}{dx} &= cos^2y \end{aligned}
dxd[tan(y)]dyd[tan(y)] dxdycos2y1 dxdydxdy=dxdx=1=1=cos2y
我们仍需要以
x
x
x 消去
y
y
y,由三角关系,
c
o
s
(
y
)
=
1
1
+
x
2
cos(y)= \frac{1}{\sqrt{1+x^2}}
cos(y)=1+x21,那么
c
o
s
2
(
y
)
=
1
1
+
x
2
cos^2(y) = \frac{1}{1+x^2}
cos2(y)=1+x21,则
d
d
x
a
r
c
t
a
n
(
x
)
=
1
1
+
x
2
.
\frac{d}{dx} arctan(x) = \frac{1}{1+x^2}.
dxdarctan(x)=1+x21.
指数函数导数公式的推导
d d x a x = a x ln a \boxed{\frac{d}{dx}a^x=a^x\ln a} dxdax=axlna
引入
由导数的定义
d
d
x
a
x
=
lim
Δ
x
→
0
a
x
+
Δ
x
−
a
x
Δ
x
=
lim
Δ
x
→
0
a
x
a
Δ
x
−
a
x
Δ
x
=
lim
Δ
x
→
0
a
x
a
Δ
x
−
1
Δ
x
=
a
x
lim
Δ
x
→
0
a
Δ
x
−
1
Δ
x
\begin{aligned} \frac{d}{dx} a^x &= \lim_{\Delta x \to 0} \frac{a^{x+\Delta x}-a^x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} \frac{a^x a^{\Delta x} - a^x}{\Delta x} \\\\ &= \lim_{\Delta x \to 0} a^x \frac{a^{\Delta x}-1}{\Delta x} \\\\ &= a^x\lim_{\Delta x \to 0} \frac{a^{\Delta x}-1}{\Delta x} \\\\ \end{aligned}
dxdax=Δx→0limΔxax+Δx−ax=Δx→0limΔxaxaΔx−ax=Δx→0limaxΔxaΔx−1=axΔx→0limΔxaΔx−1
设
lim
Δ
x
→
0
a
Δ
x
−
1
Δ
x
=
M
(
a
)
\lim_{\Delta x \to 0} \frac{a^{\Delta x}-1}{\Delta x} = M(a)
limΔx→0ΔxaΔx−1=M(a) ,则有
d
d
x
a
x
=
M
(
a
)
a
x
\frac{d}{dx}a^x = M(a)\ a^x
dxdax=M(a) ax
由于
d
d
x
a
x
∣
x
=
0
=
M
(
a
)
\frac{d}{dx} a^x |_{x=0}=M(a)
dxdax∣x=0=M(a) ,我们只需要求解出
y
=
a
x
y=a^x
y=ax 图象在
x
=
0
x=0
x=0 处切线的斜率即可得到指数函数导数公式。假设一个常数
e
e
e ,使得
M
(
e
)
=
1
M(e)=1
M(e)=1 ,那么有
d
d
x
e
x
=
e
x
\boxed{\frac{d}{dx} e^x=e^x}
dxdex=ex
同时, y = e x y=e^x y=ex 图象在 x = 0 x=0 x=0 的切线斜率为 1. 下面证 e e e 的唯一性(不完全):
随着底数
a
a
a 增长,指数函数的图象愈来愈陡峭,则其导数单调递增,当
a
=
1
a=1
a=1 时,
a
x
=
1
a^x=1
ax=1 恒成立且对应指数函数图象斜率恒为 0;我们用割线来逼近
a
=
2
a=2
a=2 以及
a
=
4
a=4
a=4 的情况:
如图,从
(
0
,
1
)
(0,1)
(0,1) 到
(
1
,
2
)
(1,2)
(1,2) 关于
y
=
2
x
y=2^x
y=2x 的割线斜率为 1,我们很容易可以看出,
M
(
2
)
<
1
M(2)<1
M(2)<1 ;
如下图,从
(
−
1
2
,
1
2
)
(-\frac{1}{2}, \frac{1}{2})
(−21,21) 到
(
1
,
0
)
(1,0)
(1,0) 关于
y
=
4
x
y=4^x
y=4x 的割线斜率为 1,我们也可以得到
M
(
4
)
>
1
M(4)>1
M(4)>1 。
由于
y
=
a
x
y=a^x
y=ax 的导数必定连续且单增,则
M
(
a
)
M(a)
M(a) 也连续且单增,由上述推理可知
a
∈
(
2
,
4
)
a \in (2, 4)
a∈(2,4) 有且只有一个
e
e
e 满足
M
(
a
)
=
1
M(a) =1
M(a)=1 ,唯一性得证。
自然对数是以
e
e
e 为底的指数函数的反函数:
y
=
e
x
,
ln
(
y
)
=
x
y=e^x, \ln (y)=x
y=ex,ln(y)=x
设
w
=
ln
x
w=\ln x
w=lnx,则有
d
d
x
e
w
=
d
d
x
x
d
d
w
(
e
w
)
d
w
d
x
=
1
e
w
d
w
d
x
=
1
d
w
d
x
=
1
e
w
\begin{aligned} \frac{d}{dx}e^w &= \frac{d}{dx} x \\\\ \frac{d}{dw}(e^w)\ \frac{dw}{dx} &= 1 \\\\ e^w\ \frac{dw}{dx} &= 1 \\\\ \frac{dw}{dx} &= \frac{1}{e^w}\\\\ \end{aligned}
dxdewdwd(ew) dxdwew dxdwdxdw=dxdx=1=1=ew1
由假设,我们有
d
d
x
ln
x
=
1
x
\boxed{\frac{d}{dx} \ln x=\frac{1}{x}}
dxdlnx=x1
证明
第一种方法:代入消元
我们使用
e
e
e 重新表达
a
x
a^x
ax :
a
x
=
(
e
ln
a
)
x
=
e
x
ln
a
a^x=\left(e^{\ln a}\right)^x=e^{x\ln a}
ax=(elna)x=exlna
由链式法则,我们可以得到
d
d
x
e
x
ln
a
=
(
ln
a
)
e
x
ln
a
.
\frac{d}{dx} e^{x \ln a}=(\ln a)\ e^{x\ln a}.
dxdexlna=(lna) exlna.
即
d
d
x
a
x
=
a
x
ln
a
\boxed{\frac{d}{dx}a^x=a^x\ln a}
dxdax=axlna
那么, M ( a ) = ln a M(a) = \ln a M(a)=lna .
第二种方法:对数微分
设 u = a x u=a^x u=ax ,而由链式法则, d d x ln u = 1 u d u d x \frac{d}{dx}\ln u=\frac{1}{u} \frac{du}{dx} dxdlnu=u1dxdu,即 ( ln u ) ′ = u ′ u . (\ln u)'=\frac{u'}{u}. (lnu)′=uu′. 则有
ln u = ln ( a x ) ln u = x ln a ( ln u ) ′ = ln a \begin{aligned} \ln u &= \ln (a^x) \\ \ln u &= x\ln a \\ (\ln u)' &= \ln a \end{aligned} lnulnu(lnu)′=ln(ax)=xlna=lna
由 ( ln u ) ′ = u ′ u (\ln u)'=\frac{u'}{u} (lnu)′=uu′ 可得 u ′ = u ( ln u ) ′ u'=u(\ln u)' u′=u(lnu)′ ,即 ( a x ) ′ = a x ln a . (a^x)' = a^x \ln a. (ax)′=axlna.
幂指函数导数公式的推导
d d x x x = x x ( 1 + ln x ) \boxed{\frac{d}{dx}x^x=x^x(1+\ln x)} dxdxx=xx(1+lnx)
之前我们已然推导出
(
ln
u
)
′
=
u
′
u
(\ln u)'=\frac{u'}{u}
(lnu)′=uu′,设
v
=
x
x
v=x^x
v=xx ,我们利用对数微分的技巧求解
v
′
v'
v′ :
ln
v
=
x
ln
x
(
ln
v
)
′
=
ln
x
+
x
⋅
1
x
=
v
′
v
\begin{aligned} \ln v &= x \ln x \\ (\ln v)' &= \ln x + x \cdot \frac{1}{x} \\ &= \frac{v'}{v} \end{aligned}
lnv(lnv)′=xlnx=lnx+x⋅x1=vv′
则有
v
′
x
x
=
1
+
ln
x
v
′
=
x
x
(
1
+
ln
x
)
.
\begin{aligned} \frac{v'}{x^x} &= 1+\ln x \\\\ v' &=x^x(1+\ln x). \end{aligned}
xxv′v′=1+lnx=xx(1+lnx).
幂法则(The Power Rule)的证明
我们先前证明过有理数幂函数的导数公式,现在我们将其拓展到实数域:
d
d
x
x
r
=
r
x
r
−
1
\boxed{\frac{d}{dx}x^r=rx^{r-1}}
dxdxr=rxr−1
第一种方法:e 的代换
显然
x
r
=
e
r
ln
x
x^r=e^{r\ln x}
xr=erlnx ,则
d
d
x
x
r
=
d
d
x
e
r
ln
x
=
e
r
ln
x
d
d
x
(
r
ln
x
)
=
e
r
ln
x
(
r
x
)
=
x
r
(
r
x
)
=
r
x
r
−
1
.
\begin{aligned} \frac{d}{dx}x^r &= \frac{d}{dx} e^{r\ln x} = e^{r\ln x}\ \frac{d}{dx}(r\ln x) \\\\ &= e^{r\ln x}\ \left(\frac{r}{x}\right) \\\\ &= x^r \left(\frac{r}{x}\right) = rx^{r-1}. \end{aligned}
dxdxr=dxderlnx=erlnx dxd(rlnx)=erlnx (xr)=xr(xr)=rxr−1.
第二种方法:对数微分
我们定义
f
(
x
)
=
x
r
f(x)=x^r
f(x)=xr,则有
ln
f
=
r
ln
x
(
ln
f
)
′
=
f
′
f
=
r
x
f
′
=
f
(
ln
f
)
′
=
x
r
(
r
x
)
=
r
x
r
−
1
.
\begin{aligned} \ln f &= r\ln x \\ (\ln f)' &= \frac{f'}{f} = \frac{r}{x} \\ f' &= f(\ln f)' = x^r \left(\frac{r}{x}\right) \\ &= rx^{r-1}. \end{aligned}
lnf(lnf)′f′=rlnx=ff′=xr=f(lnf)′=xr(xr)=rxr−1.
e 的极限本质
lim n → ∞ ( 1 + 1 n ) n = e \boxed{\lim_{n\to \infty}\left(1+\frac{1}{n}\right)^n = e} n→∞lim(1+n1)n=e
取自然对数
ln
[
(
1
+
1
n
)
n
]
=
n
ln
(
1
+
1
n
)
\ln \left[\left(1+\frac{1}{n}\right)^n\right] = n\ln \left(1+\frac{1}{n}\right)
ln[(1+n1)n]=nln(1+n1)
当 n → ∞ n\to \infty n→∞ 时, Δ x = 1 n → 0 \Delta x =\frac{1}{n} \to 0 Δx=n1→0, n = 1 Δ x n=\frac{1}{\Delta x} n=Δx1,这样我们就可以将其转换为趋近于 0 的极限问题:
lim n → ∞ n ln ( 1 + 1 n ) = lim Δ x → 0 [ 1 Δ x ln ( 1 + Δ x ) ] \begin{aligned} \lim_{n\to \infty}n\ln \left(1+\frac{1}{n}\right) &= \lim_{\Delta x \to 0} \left[\frac{1}{\Delta x} \ln (1+\Delta x)\right] \\\\ \end{aligned} n→∞limnln(1+n1)=Δx→0lim[Δx1ln(1+Δx)]
利用
ln
1
=
0
\ln 1 = 0
ln1=0 这一性质,我们可以构造出熟悉的导数极限定义式的形式:
lim
Δ
x
→
0
[
1
Δ
x
ln
(
1
+
Δ
x
)
]
=
lim
Δ
x
→
0
{
1
Δ
x
[
ln
(
1
+
Δ
x
)
−
ln
1
]
}
=
lim
Δ
x
→
0
ln
(
1
+
Δ
x
)
−
ln
1
Δ
x
=
d
d
x
ln
x
∣
x
=
1
=
1
x
∣
x
=
1
=
1
\begin{aligned} \lim_{\Delta x \to 0} \left[\frac{1}{\Delta x} \ln (1+\Delta x)\right] &= \lim_{\Delta x\to 0} \left\{ \frac{1}{\Delta x} \big [ \ln (1+\Delta x)-\ln 1 \big ]\right\} \\\\ &= \lim_{\Delta x \to 0} \frac{\ln (1+\Delta x)-\ln 1}{\Delta x} \\\\ &= \frac{d}{dx} \ln x\ \bigg|_{x=1} = \frac{1}{x}\ \bigg|_{x=1} \\\\ &= 1 \end{aligned}
Δx→0lim[Δx1ln(1+Δx)]=Δx→0lim{Δx1[ln(1+Δx)−ln1]}=Δx→0limΔxln(1+Δx)−ln1=dxdlnx
x=1=x1
x=1=1
这是原极限取自然对数的结果,意味着
lim
n
→
∞
(
1
+
1
n
)
n
=
lim
n
→
∞
e
ln
[
(
1
+
1
n
)
n
]
=
e
lim
n
→
∞
ln
[
(
1
+
1
n
)
n
]
=
e
1
=
e
.
\begin{aligned} \lim_{n\to \infty}\left(1+\frac{1}{n}\right)^n &= \lim_{n\to \infty} e^{\ln\left[ \left(1+\frac{1}{n}\right)^n\right]} \\\\ &= e^{\lim_{n\to \infty}\ln\left[ \left(1+\frac{1}{n}\right)^n\right]} \\\\ &= e^1 = e. \end{aligned}
n→∞lim(1+n1)n=n→∞limeln[(1+n1)n]=elimn→∞ln[(1+n1)n]=e1=e.
双曲函数导数公式的推导
双曲正弦函数
sinh
(
x
)
=
e
x
−
e
−
x
2
\sinh(x) = \frac{e^x-e^{-x}}{2}
sinh(x)=2ex−e−x
双曲余弦函数
cosh
(
x
)
=
e
x
+
e
−
x
2
\cosh(x) = \frac{e^x+e^{-x}}{2}
cosh(x)=2ex+e−x
则有
d
d
x
sinh
(
x
)
=
cosh
(
x
)
\boxed{\frac{d}{dx}\sinh(x) = \cosh(x)}
dxdsinh(x)=cosh(x)
d d x cosh ( x ) = sinh ( x ) \boxed{\frac{d}{dx}\cosh(x) = \sinh(x)} dxdcosh(x)=sinh(x)
d d x sinh ( x ) = e x − ( − e − x ) 2 = cosh ( x ) . \frac{d}{dx} \sinh(x) = \frac{e^x-(-e^{-x})}{2} = \cosh(x). dxdsinh(x)=2ex−(−e−x)=cosh(x).
cosh ( x ) \cosh(x) cosh(x) 同理可证。
重要性质的证明
cosh 2 ( x ) − sinh 2 ( x ) = 1 \boxed{\cosh^2(x)-\sinh^2(x)=1} cosh2(x)−sinh2(x)=1
cosh 2 ( x ) − sinh 2 ( x ) = ( e x + e − x 2 ) 2 − ( e x − e − x 2 ) 2 = 1 4 ( e 2 x + 2 e x e − x + e − 2 x ) − 1 4 ( e 2 x − 2 + e − 2 x ) = 1 4 ( 2 + 2 ) = 1. \begin{aligned} \cosh^2(x)-\sinh^2(x) &= \left(\frac{e^x+e^{-x}}{2}\right)^2-\left(\frac{e^x-e^{-x}}{2}\right)^2 \\\\ &= \frac{1}{4} (e^{2x}+2e^xe^{-x}+e^{-2x}) - \frac{1}{4} (e^{2x}-2+e^{-2x}) \\\\ &= \frac{1}{4} (2+2) = 1. \end{aligned} cosh2(x)−sinh2(x)=(2ex+e−x)2−(2ex−e−x)2=41(e2x+2exe−x+e−2x)−41(e2x−2+e−2x)=41(2+2)=1.