偏导数与全微分
偏导数与全微分的概念
现在,我们把导数和微分的概念,推广到多元函数的情形。只不过,在二维以上,函数的方向十分复杂,绝不只有左导数和右导数两个方向。然而,我们可以先对某个变元求导数,称为偏导数。
定义14.1(偏导数)
f
(
x
1
,
x
2
,
⋯
,
x
n
)
f(x_1,x_2,\cdots,x_n)
f(x1,x2,⋯,xn)在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)的某个邻域上有定义,如果对第
i
(
1
≤
i
≤
n
)
i(1\le i \le n)
i(1≤i≤n)的变元,极限
lim
Δ
x
i
→
0
f
(
x
1
0
,
⋯
,
x
i
−
1
0
,
x
i
,
x
i
+
1
0
,
⋯
,
x
n
0
)
−
f
(
x
1
0
,
⋯
,
x
n
0
)
Δ
x
i
\lim_{\Delta x_i\to 0}{\frac{f(x_1^0,\cdots,x_{i-1}^0,x_i,x_{i+1}^0,\cdots,x_n^0)-f(x_1^0,\cdots,x_n^0)}{\Delta x_i}}
Δxi→0limΔxif(x10,⋯,xi−10,xi,xi+10,⋯,xn0)−f(x10,⋯,xn0)存在,称该极限为
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处对
x
i
x_i
xi的偏导数,\记为
∂
f
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
i
\frac{\partial f(x_1^0,\cdots,x_n^0)}{\partial x_i}
∂xi∂f(x10,⋯,xn0)或
f
i
(
x
1
0
,
⋯
,
x
n
0
)
f_i(x_1^0,\cdots,x_n^0)
fi(x10,⋯,xn0)
如果
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)在某个开集
E
E
E上每个点对所有变元的偏导数都存在,那么,对各个变元的偏导数,都是这个开集上的一个
n
n
n元函数,同样可以讨论极限、连续性的等概念。
我们再一元函数上还有微分的概念,在一元函数上,全微分定义成某点的"切线",在二元函数上,全微分就应该是某点的切平面,在三维以上,就是切“超平面”,只不过,这时我们没有几何直观可以参考。
一维上的直线可以表为
y
=
a
+
b
x
y=a+bx
y=a+bx
二维上的平面可表为
y
=
a
+
b
1
x
+
b
2
x
y=a+b_1x+b_2x
y=a+b1x+b2x
推广到
n
n
n维上,超平面可表为
y
=
a
+
∑
k
=
1
n
b
k
x
k
y=a+\sum_{k=1}^{n}{b_kx_k}
y=a+∑k=1nbkxk
所谓全微分,就是在函数在某点附近,可以用一个超平面近似,即:
f
(
x
)
=
f
(
x
0
)
+
∑
k
=
1
n
b
k
Δ
x
k
+
o
(
∣
∣
Δ
x
∣
∣
)
f(x)=f(x_0)+\sum_{k=1}^n{b_k\Delta x_k}+o(||\Delta x||)
f(x)=f(x0)+k=1∑nbkΔxk+o(∣∣Δx∣∣)
定义14.2(全微分)
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)在
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
x_0=(x_1^0,\cdots,x_n^0)
x0=(x10,⋯,xn0)的某个邻域上有定义,如果
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)可表为
f
(
x
1
,
⋯
,
x
n
)
=
f
(
x
1
0
,
⋯
,
x
n
0
)
+
∑
k
=
1
n
A
k
(
x
k
−
x
k
0
)
+
o
(
∣
∣
x
−
x
0
∣
∣
)
f(x_1,\cdots,x_n)=f(x_1^0,\cdots,x_n^0)+\sum_{k=1}^n{A_k(x_k-x_k^0)}+ o(||x-x_0||)
f(x1,⋯,xn)=f(x10,⋯,xn0)+k=1∑nAk(xk−xk0)+o(∣∣x−x0∣∣)其中
A
1
,
⋯
,
A
n
A_1,\cdots,A_n
A1,⋯,An和
Δ
x
=
x
−
x
0
\Delta x = x-x_0
Δx=x−x0无关,则称
f
(
x
)
f(x)
f(x)在
x
0
x_0
x0处可微,超平面
∑
k
=
1
n
A
k
d
x
k
\sum_{k=1}^n{A_kdx_k}
∑k=1nAkdxk称为
f
(
x
)
f(x)
f(x)在
x
0
x_0
x0处的全微分,记为
d
f
=
∑
k
=
1
n
A
k
d
x
k
df = \sum_{k=1}^n{A_kdx_k}
df=∑k=1nAkdxk
定理14.1(可微的必要条件)
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)在
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
x_0=(x_1^0,\cdots,x_n^0)
x0=(x10,⋯,xn0)的某个邻域上有定义,
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)在
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
x_0=(x_1^0,\cdots,x_n^0)
x0=(x10,⋯,xn0)上可微,则
f
f
f在
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
x_0=(x_1^0,\cdots,x_n^0)
x0=(x10,⋯,xn0)对各变元可求偏导,并且:
d
f
=
∑
k
=
1
n
f
k
(
x
1
0
,
⋯
,
x
n
0
)
d
x
k
df = \sum_{k=1}^n{f_k(x_1^0,\cdots,x_n^0)dx_k}
df=k=1∑nfk(x10,⋯,xn0)dxk
这由全微分的定义可以直接验证。其次,容易验证可微必连续。但就算n元函数在某点对各变元可求偏导且连续,也不一定可微。
例14.1 f ( x , y ) = { x y x 2 + y 2 x 2 + y 2 > 0 0 x = 0 , y = 0 f(x,y)= \begin{cases} \frac{xy}{\sqrt{x^2+y^2}}&x^2+y^2>0\\ 0&x=0,y=0 \end{cases} f(x,y)={x2+y2xy0x2+y2>0x=0,y=0在 f ( x , y ) f(x,y) f(x,y)在 ( 0 , 0 ) (0,0) (0,0)处连续且对各变元可求偏导,然而: lim ( x , y ) → ( 0 , 0 ) f ( x , y ) x 2 + y 2 = lim ( x , y ) → ( 0 , 0 ) x y x 2 + y 2 \lim_{(x,y)\to(0,0)}{\frac{f(x,y)}{\sqrt{x^2+y^2}}} =\lim_{(x,y)\to(0,0)}{\frac{xy}{x^2+y^2}} (x,y)→(0,0)limx2+y2f(x,y)=(x,y)→(0,0)limx2+y2xy极限不存在,因此, f ( x , y ) f(x,y) f(x,y)在 ( 0 , 0 ) (0,0) (0,0)点不可微
那么,满足何种条件能够可微呢?下面我们给出一个充分条件:
定理14.2(可微的充分条件) f ( x 1 , ⋯ , x n ) f(x_1,\cdots,x_n) f(x1,⋯,xn)在 x 0 = ( x 1 0 , ⋯ , x n 0 ) x_0=(x_1^0,\cdots,x_n^0) x0=(x10,⋯,xn0)的某个邻域上有定义且对各变元可求偏导,并且各偏导在 x 0 x_0 x0处连续,则 f ( x 1 , ⋯ , x n ) f(x_1,\cdots,x_n) f(x1,⋯,xn)在 x 0 x_0 x0处可微
证:
f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) = f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) + ∑ k = 1 n − 1 ( − f ( x 1 0 , ⋯ , x k 0 , x k + 1 , ⋯ , x n ) + f ( x 1 0 , ⋯ , x k 0 , x k + 1 , ⋯ , x n ) ) = ∑ k = 1 n [ f ( x 1 0 , ⋯ , x k − 1 0 , x k , x k + 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x k 0 , x k + 1 , ⋯ , x n ) ] f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0) =f(x_1,\cdots,x_n)-\\f(x_1^0,\cdots,x_n^0)+\sum_{k=1}^{n-1}{( -f(x_1^0,\cdots,x_k^0,x_{k+1},\cdots,x_n) +f(x_1^0,\cdots,x_k^0,x_{k+1},\cdots,x_n))}\\ =\sum_{k=1}^n{[f(x_1^0,\cdots,x_{k-1}^0,x_k,x_{k+1},\cdots,x_n) -f(x_1^0,\cdots,x_k^0,x_{k+1},\cdots,x_n)]} f(x1,⋯,xn)−f(x10,⋯,xn0)=f(x1,⋯,xn)−f(x10,⋯,xn0)+k=1∑n−1(−f(x10,⋯,xk0,xk+1,⋯,xn)+f(x10,⋯,xk0,xk+1,⋯,xn))=k=1∑n[f(x10,⋯,xk−10,xk,xk+1,⋯,xn)−f(x10,⋯,xk0,xk+1,⋯,xn)] 由拉格朗日中值定理,存在 ξ k \xi_k ξk介于 x k x_k xk和 x k 0 x_k^0 xk0之间 f ( x 1 , ⋯ , x n ) = f ( x 1 0 , ⋯ , x n 0 ) + ∑ k = 1 n f k ( x 1 0 , ⋯ , x k − 1 0 , ξ k , x k + 1 , ⋯ , x n ) Δ x k = ∑ k = 1 n f k ( x 1 0 , ⋯ , x n 0 ) Δ x k + ∑ k = 1 n [ f k ( x 1 0 , ⋯ , x k − 1 0 , ξ k , x k + 1 , ⋯ , x n ) − f k ( x 1 0 , ⋯ , x n 0 ) ] Δ x k f(x_1,\cdots,x_n)=f(x_1^0,\cdots,x_n^0) +\sum_{k=1}^n{f_k(x_1^0,\cdots,x_{k-1}^0,\xi_k,x_{k+1},\cdots,x_n)\Delta x_k}\\ =\sum_{k=1}^n{f_k(x_1^0,\cdots,x_n^0)\Delta x_k} +\sum_{k=1}^n[f_k(x_1^0,\cdots,x_{k-1}^0,\xi_k,x_{k+1},\cdots,x_n)-f_k(x_1^0,\cdots,x_n^0)]\Delta x_k f(x1,⋯,xn)=f(x10,⋯,xn0)+k=1∑nfk(x10,⋯,xk−10,ξk,xk+1,⋯,xn)Δxk=k=1∑nfk(x10,⋯,xn0)Δxk+k=1∑n[fk(x10,⋯,xk−10,ξk,xk+1,⋯,xn)−fk(x10,⋯,xn0)]Δxk考察余项: f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) − ∑ k = 1 n f k ( x 1 0 , ⋯ , x n 0 ) Δ x k ∑ k = 1 n Δ 2 x k = ∑ k = 1 n [ f k ( x 1 0 , ⋯ , x k − 1 0 , ξ k , x k + 1 , ⋯ , x n ) − f k ( x 1 0 , ⋯ , x n 0 ) ] Δ x k ∑ i = 1 n Δ 2 x i \frac{f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)-\sum_{k=1}^n{f_k(x_1^0,\cdots,x_n^0)\Delta x_k}}{\sqrt{\sum_{k=1}^n{\Delta^2 x_k}}}\\ =\sum_{k=1}^n[f_k(x_1^0,\cdots,x_{k-1}^0,\xi_k,x_{k+1},\cdots,x_n)-f_k(x_1^0,\cdots,x_n^0)]\frac{\Delta x_k}{\sqrt{\sum_{i=1}^n\Delta^2 x_i}} ∑k=1nΔ2xkf(x1,⋯,xn)−f(x10,⋯,xn0)−∑k=1nfk(x10,⋯,xn0)Δxk=k=1∑n[fk(x10,⋯,xk−10,ξk,xk+1,⋯,xn)−fk(x10,⋯,xn0)]∑i=1nΔ2xiΔxk ∣ Δ x k ∑ i = 1 n Δ 2 x i ∣ ≤ 1 |\frac{\Delta x_k}{\sqrt{\sum_{i=1}^n\Delta^2 x_i}}|\le 1 ∣∑i=1nΔ2xiΔxk∣≤1因此: ∣ f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) − ∑ k = 1 n f k ( x 1 0 , ⋯ , x n 0 ) Δ x k ∑ k = 1 n Δ 2 x k ∣ ≤ ∑ k = 1 n ∣ f k ( x 1 0 , ⋯ , x k − 1 0 , ξ k , x k + 1 , ⋯ , x n ) − f k ( x 1 0 , ⋯ , x n 0 ) ∣ |\frac{f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)-\sum_{k=1}^n{f_k(x_1^0,\cdots,x_n^0)\Delta x_k}}{\sqrt{\sum_{k=1}^n{\Delta^2 x_k}}}|\\ \le \sum_{k=1}^n|f_k(x_1^0,\cdots,x_{k-1}^0,\xi_k,x_{k+1},\cdots,x_n)-f_k(x_1^0,\cdots,x_n^0)| ∣∑k=1nΔ2xkf(x1,⋯,xn)−f(x10,⋯,xn0)−∑k=1nfk(x10,⋯,xn0)Δxk∣≤k=1∑n∣fk(x10,⋯,xk−10,ξk,xk+1,⋯,xn)−fk(x10,⋯,xn0)∣再由偏导数的连续性,就有 lim ( x 1 , ⋯ , x n ) → ( x 1 0 , ⋯ , x n 0 ) ∣ f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) − ∑ k = 1 n f k ( x 1 0 , ⋯ , x n 0 ) Δ x k ∑ k = 1 n Δ 2 x k ∣ = 0 \lim_{(x_1,\cdots,x_n)\to(x_1^0,\cdots,x_n^0)}{|\frac{f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)-\sum_{k=1}^n{f_k(x_1^0,\cdots,x_n^0)\Delta x_k}}{\sqrt{\sum_{k=1}^n{\Delta^2 x_k}}}|}=0 (x1,⋯,xn)→(x10,⋯,xn0)lim∣∑k=1nΔ2xkf(x1,⋯,xn)−f(x10,⋯,xn0)−∑k=1nfk(x10,⋯,xn0)Δxk∣=0因此, f ( x 1 , ⋯ , x n ) f(x_1,\cdots,x_n) f(x1,⋯,xn)在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处可微
我们把偏导数连续称为连续可微。这样,可微、可导和连续性的关系可以概括为:
(1)连续可微一定可微
(2)可微一定可求偏导数
(3)可微一定连续
(4)连续不一定可求偏导
(5)可求偏导不一定可微
多元函数微分法则
为了给出多元情形下的求导和微分法则,我们首先给出向量值函数的全微分概念
定义14.3 g ( x 1 , x 2 , ⋯ , x n ) = ( g 1 ( x 1 , ⋯ , x n ) , ⋯ , g m ( x 1 , ⋯ , x n ) ) g(x_1,x_2,\cdots,x_n)=(g_1(x_1,\cdots,x_n),\cdots,g_m(x_1,\cdots,x_n)) g(x1,x2,⋯,xn)=(g1(x1,⋯,xn),⋯,gm(x1,⋯,xn))是 n n n元 m m m维向量值函数,在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)附近有定义,如果存在与 ( x 1 , ⋯ , x n ) (x_1,\cdots,x_n) (x1,⋯,xn)无关,仅与 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)有关的 m m m行 n n n列矩阵 A A A,记 Δ x = ( x 1 − x 1 0 , ⋯ , x n − x n 0 ) T \Delta x = (x_1-x_1^0,\cdots,x_n-x_n^0)^T Δx=(x1−x10,⋯,xn−xn0)T,使得 [ g 1 ( x 1 , ⋯ , x n ) − g 1 ( x 1 0 , ⋯ , x n 0 ) ⋯ g m ( x 1 , ⋯ , x n ) − g m ( x 1 0 , ⋯ , x n 0 ) ] = A Δ x + [ o 1 ( ∣ ∣ Δ x ∣ ∣ ) ⋯ o m ( ∣ ∣ Δ x ∣ ∣ ) ] \left[\begin{matrix} g_1(x_1,\cdots,x_n)-g_1(x_1^0,\cdots,x_n^0)\\ \cdots\\ g_m(x_1,\cdots,x_n)-g_m(x_1^0,\cdots,x_n^0) \end{matrix}\right] =A\Delta x + \left[ \begin{matrix} o_1(||\Delta x||)\\ \cdots\\ o_m(||\Delta x||) \end{matrix} \right] ⎣⎡g1(x1,⋯,xn)−g1(x10,⋯,xn0)⋯gm(x1,⋯,xn)−gm(x10,⋯,xn0)⎦⎤=AΔx+⎣⎡o1(∣∣Δx∣∣)⋯om(∣∣Δx∣∣)⎦⎤则称向量值函数 g g g在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处可微,矩阵 A A A称为 g g g在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处的Frechet导数。 A d x Adx Adx称为 g g g在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处的全微分。
实际上,由定义容易得出,如果向量值函数在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微的充要条件是每个分量函数都在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,并且,Frechet导数就等于:
[
∂
g
1
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
1
⋯
∂
g
1
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
n
⋯
⋯
⋯
∂
g
m
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
1
⋯
∂
g
m
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
n
]
\left[\begin{matrix} \frac{\partial g_1(x_1^0,\cdots,x_n^0)}{\partial x_1}&\cdots&\frac{\partial g_1(x_1^0,\cdots,x_n^0)}{\partial x_n}\\ \cdots&\cdots&\cdots\\ \frac{\partial g_m(x_1^0,\cdots,x_n^0)}{\partial x_1}&\cdots&\frac{\partial g_m(x_1^0,\cdots,x_n^0)}{\partial x_n} \end{matrix}\right]
⎣⎢⎡∂x1∂g1(x10,⋯,xn0)⋯∂x1∂gm(x10,⋯,xn0)⋯⋯⋯∂xn∂g1(x10,⋯,xn0)⋯∂xn∂gm(x10,⋯,xn0)⎦⎥⎤为了方便,我们把Frechet导数记为
g
′
(
x
0
)
g^\prime(x_0)
g′(x0),对
n
n
n元函数来说,Frechet导数就是
n
n
n维的行向量。实际上,Frechet导数就是一元导数的的一个推广,Frechet可导就等价于可微,在这层意义下,可微和可导是等价的。
定理14.3(线性性质)
g
1
,
g
2
g_1,g_2
g1,g2是
n
n
n元
m
m
m维向量值函数并且都在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,则
(1)
g
1
+
g
2
g_1+g_2
g1+g2在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,并且
g
1
′
(
x
1
0
,
⋯
,
x
n
0
)
+
g
2
′
(
x
1
0
,
⋯
,
x
n
0
)
=
(
g
1
+
g
2
)
′
(
x
1
0
,
⋯
,
x
n
0
)
g_1^\prime(x_1^0,\cdots,x_n^0)+g_2^\prime(x_1^0,\cdots,x_n^0) =(g_1+g_2)^\prime(x_1^0,\cdots,x_n^0)
g1′(x10,⋯,xn0)+g2′(x10,⋯,xn0)=(g1+g2)′(x10,⋯,xn0)(2)
c
c
c是任意实数,
c
g
1
cg_1
cg1在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,并且
c
g
1
′
(
x
1
0
,
⋯
,
x
n
0
)
=
(
c
g
1
)
′
(
x
1
0
,
⋯
,
x
n
0
)
cg_1^\prime(x_1^0,\cdots,x_n^0) = (cg_1)^\prime(x_1^0,\cdots,x_n^0)
cg1′(x10,⋯,xn0)=(cg1)′(x10,⋯,xn0)
这两个性质按照Frechet导数的定义是显然的。
定理14.4
g
g
g是
n
n
n元
m
m
m维向量值函数,并且在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,
f
f
f是
n
n
n元函数,并且在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,则
F
=
g
(
x
1
,
⋯
,
x
n
)
f
(
x
1
,
⋯
,
x
n
)
F=g(x_1,\cdots,x_n)f(x_1,\cdots,x_n)
F=g(x1,⋯,xn)f(x1,⋯,xn)在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,并且
F
′
(
x
1
0
,
⋯
,
x
n
0
)
=
g
(
x
1
0
,
⋯
,
x
n
0
)
f
′
(
x
1
0
,
⋯
,
x
n
0
)
+
g
′
(
x
1
0
,
⋯
,
x
n
0
)
f
(
x
1
0
,
⋯
,
x
n
0
)
F^\prime(x_1^0,\cdots,x_n^0)=g(x_1^0,\cdots,x_n^0)f^\prime(x_1^0,\cdots,x_n^0) +g^\prime(x_1^0,\cdots,x_n^0)f(x_1^0,\cdots,x_n^0)
F′(x10,⋯,xn0)=g(x10,⋯,xn0)f′(x10,⋯,xn0)+g′(x10,⋯,xn0)f(x10,⋯,xn0)
在理解定理14.4时,需要注意的是 f g fg fg是 n n n元 m m m维向量值函数,其Frechet导数是 m m m行 n n n列矩阵,等式右边第一项中: g g g是 m m m行的列向量, f ′ f^\prime f′是 n n n列的行向量,而第二项是一个数乘的形式。通过定理14.4,多元导数就和一元导数在乘法运算法则上统一起来了。下面证明定理14.4
证:
首先 g ( x 1 , ⋯ , x n ) f ( x 1 , ⋯ , x n ) − g ( x 1 0 , ⋯ , x n 0 ) f ( x 1 0 , ⋯ , x n 0 ) = g ( x 1 , ⋯ , x n ) f ( x 1 , ⋯ , x n ) − g ( x 1 0 , ⋯ , x n 0 ) f ( x 1 , ⋯ , x n ) + g ( x 1 0 , ⋯ , x n 0 ) f ( x 1 , ⋯ , x n ) − g ( x 1 0 , ⋯ , x n 0 ) f ( x 1 0 , ⋯ , x n 0 ) g(x_1,\cdots,x_n)f(x_1,\cdots,x_n)-g(x_1^0,\cdots,x_n^0) f(x_1^0,\cdots,x_n^0)\\ =g(x_1,\cdots,x_n)f(x_1,\cdots,x_n)-g(x_1^0,\cdots,x_n^0)f(x_1,\cdots,x_n)\\ +g(x_1^0,\cdots,x_n^0)f(x_1,\cdots,x_n)-g(x_1^0,\cdots,x_n^0) f(x_1^0,\cdots,x_n^0) g(x1,⋯,xn)f(x1,⋯,xn)−g(x10,⋯,xn0)f(x10,⋯,xn0)=g(x1,⋯,xn)f(x1,⋯,xn)−g(x10,⋯,xn0)f(x1,⋯,xn)+g(x10,⋯,xn0)f(x1,⋯,xn)−g(x10,⋯,xn0)f(x10,⋯,xn0)其次,由可微性,就有 g ( x 1 , ⋯ , x n ) − g ( x 1 0 , ⋯ , x n 0 ) = g ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 1 ( ∣ ∣ Δ x ∣ ∣ ) g(x_1,\cdots,x_n)-g(x_1^0,\cdots,x_n^0)= g^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_1(||\Delta x||) g(x1,⋯,xn)−g(x10,⋯,xn0)=g′(x10,⋯,xn0)Δx+o1(∣∣Δx∣∣) f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) = f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)= f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||) f(x1,⋯,xn)−f(x10,⋯,xn0)=f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)再令 h ( x 1 , ⋯ , x n ) = f ( x 1 , ⋯ , x n ) o 1 ( ∣ ∣ Δ x ∣ ∣ ) + [ f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) ] Δ x + g ( x 1 0 , ⋯ , x n 0 ) o 2 ( ∣ ∣ Δ x ∣ ∣ ) h(x_1,\cdots,x_n)=f(x_1,\cdots,x_n)o_1(||\Delta x||) \\+[f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)]\Delta x +g(x_1^0,\cdots,x_n^0)o_2(||\Delta x||) h(x1,⋯,xn)=f(x1,⋯,xn)o1(∣∣Δx∣∣)+[f(x1,⋯,xn)−f(x10,⋯,xn0)]Δx+g(x10,⋯,xn0)o2(∣∣Δx∣∣)由于 ∣ Δ x k ∣ ∣ Δ x ∣ ∣ ∣ ≤ 1 |\frac{\Delta x_k}{||\Delta x||}|\le 1 ∣∣∣Δx∣∣Δxk∣≤1再由 f ( x 1 , ⋯ , x n ) f(x_1,\cdots,x_n) f(x1,⋯,xn)
在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处的连续性,就有 lim ∣ ∣ Δ x ∣ ∣ → 0 [ f ( x 1 , ⋯ , x n ) − f ( x 1 0 , ⋯ , x n 0 ) ] Δ x ∣ ∣ Δ x ∣ ∣ = 0 \lim_{||\Delta x|| \to 0}{\frac{[f(x_1,\cdots,x_n)-f(x_1^0,\cdots,x_n^0)]\Delta x}{||\Delta x||}}=0 ∣∣Δx∣∣→0lim∣∣Δx∣∣[f(x1,⋯,xn)−f(x10,⋯,xn0)]Δx=0因此 lim ∣ ∣ Δ x ∣ ∣ → 0 h ( x 1 , ⋯ , x n ) ∣ ∣ Δ x ∣ ∣ = 0 \lim_{||\Delta x|| \to 0}{\frac{h(x_1,\cdots,x_n)}{||\Delta x||}}=0 ∣∣Δx∣∣→0lim∣∣Δx∣∣h(x1,⋯,xn)=0而 F ( x 1 , ⋯ , x n ) − F ( x 1 0 , ⋯ , x n 0 ) = [ g ( x 1 0 , ⋯ , x n 0 ) f ′ ( x 1 0 , ⋯ , x n 0 ) + g ′ ( x 1 0 , ⋯ , x n 0 ) f ( x 1 0 , ⋯ , x n 0 ) ] Δ x + h ( x 1 , ⋯ , x n ) F(x_1,\cdots,x_n)-F(x_1^0,\cdots,x_n^0)=[g(x_1^0,\cdots,x_n^0)f^\prime(x_1^0,\cdots,x_n^0) \\+g^\prime(x_1^0,\cdots,x_n^0)f(x_1^0,\cdots,x_n^0)]\Delta x + h(x_1,\cdots,x_n) F(x1,⋯,xn)−F(x10,⋯,xn0)=[g(x10,⋯,xn0)f′(x10,⋯,xn0)+g′(x10,⋯,xn0)f(x10,⋯,xn0)]Δx+h(x1,⋯,xn)
接下来我们给出多元下的复合函数求导法则:
定理14.5
f
f
f是
n
n
n元
m
m
m维向量函数,在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可导,
(
y
1
0
,
⋯
,
y
m
0
)
=
f
(
x
1
0
,
⋯
,
x
n
0
)
(y_1^0,\cdots,y_m^0)=f(x_1^0,\cdots,x_n^0)
(y10,⋯,ym0)=f(x10,⋯,xn0),
g
g
g是
m
m
m元
k
k
k维向量函数,在
(
y
1
0
,
⋯
,
y
m
0
)
(y_1^0,\cdots,y_m^0)
(y10,⋯,ym0)处可导,则
f
(
g
(
x
1
,
⋯
,
x
n
)
)
f(g(x_1,\cdots,x_n))
f(g(x1,⋯,xn))在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可导,并且
(
g
(
f
(
x
1
0
,
⋯
,
x
n
0
)
)
)
′
=
g
′
(
f
(
x
1
0
,
⋯
,
x
n
0
)
)
f
′
(
x
1
0
,
⋯
,
x
n
0
)
(g(f(x_1^0,\cdots,x_n^0)))^\prime = g^\prime(f(x_1^0,\cdots,x_n^0))f^\prime(x_1^0,\cdots,x_n^0)
(g(f(x10,⋯,xn0)))′=g′(f(x10,⋯,xn0))f′(x10,⋯,xn0)
证:
g g g在 ( y 1 0 , ⋯ , y m 0 ) (y_1^0,\cdots,y_m^0) (y10,⋯,ym0)可微,则 g ( y 1 , ⋯ , y m ) − g ( y 1 0 , ⋯ , y m 0 ) = g ′ ( y 1 0 , ⋯ , y m 0 ) Δ y + o 1 ( ∣ ∣ Δ y ∣ ∣ ) (1) \tag{1} g(y_1,\cdots,y_m)-g(y_1^0,\cdots,y_m^0)=\\ g^\prime(y_1^0,\cdots,y_m^0)\Delta y + o_1(||\Delta y||) g(y1,⋯,ym)−g(y10,⋯,ym0)=g′(y10,⋯,ym0)Δy+o1(∣∣Δy∣∣)(1)再由 f f f在 ( x 1 0 , ⋯ , x n 0 ) (x_1^0,\cdots,x_n^0) (x10,⋯,xn0)处可微,则 f ( x 1 , ⋯ , x n ) − ( y 1 0 , ⋯ , y m 0 ) = f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) (2) \tag{2} f(x_1,\cdots,x_n)-(y_1^0,\cdots,y_m^0)= f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||) f(x1,⋯,xn)−(y10,⋯,ym0)=f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)(2)而 ∣ ∣ f ( x 1 , ⋯ , x n ) − ( y 1 , ⋯ , y m ) ∣ ∣ ∣ ∣ Δ x ∣ ∣ = ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ ∣ ∣ \frac{||f(x_1,\cdots,x_n)-(y_1,\cdots,y_m)||}{||\Delta x||} =||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||)}{||\Delta x||}|| ∣∣Δx∣∣∣∣f(x1,⋯,xn)−(y1,⋯,ym)∣∣=∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)∣∣由范数的性质,有 ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ ∣ ∣ ≤ ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x ∣ ∣ Δ x ∣ ∣ ∣ ∣ + ∣ ∣ o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ ∣ ∣ ||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||)}{||\Delta x||}|| \le ||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x}{||\Delta x||}||+||\frac{o_2(||\Delta x||)}{||\Delta x||}|| ∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)∣∣≤∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx∣∣+∣∣∣∣Δx∣∣o2(∣∣Δx∣∣)∣∣而 lim ∣ ∣ Δ x ∣ ∣ → 0 o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ = 0 \lim_{||\Delta x||\to 0}{\frac{o_2(||\Delta x||)}{||\Delta x||}} = 0 ∣∣Δx∣∣→0lim∣∣Δx∣∣o2(∣∣Δx∣∣)=0同时设 f i j ( x 1 0 , ⋯ , x n 0 ) f_{ij}(x_1^0,\cdots,x_n^0) fij(x10,⋯,xn0)是 f f f的第 i i i个分量对第 j j j个变元的偏导数,则 f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x = [ ∑ i = 1 n f 1 i ( x 1 0 , ⋯ , x n 0 ) Δ x i ⋯ ∑ i = 1 n f m i ( x 1 0 , ⋯ , x n 0 ) Δ x i ] f^\prime(x_1^0,\cdots,x_n^0)\Delta x= \left[ \begin{matrix} \sum_{i=1}^n{f_{1i}(x_1^0,\cdots,x_n^0)\Delta x_i}\\ \cdots\\ \sum_{i=1}^n{f_{mi}(x_1^0,\cdots,x_n^0)\Delta x_i} \end{matrix} \right] f′(x10,⋯,xn0)Δx=⎣⎡∑i=1nf1i(x10,⋯,xn0)Δxi⋯∑i=1nfmi(x10,⋯,xn0)Δxi⎦⎤对任意的 1 ≤ i ≤ m 1\le i \le m 1≤i≤m,都要 ∣ ∑ j − 1 n f i j ( x 1 0 , ⋯ , x n 0 ) Δ x j ∣ ≤ ∑ j = 1 n f i j 2 ( x 1 0 , ⋯ , x n 0 ) ∣ ∣ Δ x ∣ ∣ |\sum_{j-1}^n f_{ij}(x_1^0,\cdots,x_n^0)\Delta x_j| \le \sqrt{\sum_{j=1}^n{f_{ij}^2(x_1^0,\cdots,x_n^0)}} ||\Delta x|| ∣j−1∑nfij(x10,⋯,xn0)Δxj∣≤j=1∑nfij2(x10,⋯,xn0)∣∣Δx∣∣因此 ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x ∣ ∣ ≤ ∑ i = 1 m ∑ j = 1 n f i j 2 ( x 1 0 , ⋯ , x n 0 ) ∣ ∣ Δ x ∣ ∣ ||f^\prime(x_1^0,\cdots,x_n^0)\Delta x|| \le \sqrt{\sum_{i=1}^{m}\sum_{j=1}^{n}{f_{ij}^2(x_1^0,\cdots,x_n^0)}}||\Delta x|| ∣∣f′(x10,⋯,xn0)Δx∣∣≤i=1∑mj=1∑nfij2(x10,⋯,xn0)∣∣Δx∣∣ ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x ∣ ∣ Δ x ∣ ∣ ∣ ∣ ≤ ∑ i = 1 m ∑ j = 1 n f i j 2 ( x 1 0 , ⋯ , x n 0 ) ||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x}{||\Delta x||}|| \le \sqrt{\sum_{i=1}^{m}\sum_{j=1}^{n}{f_{ij}^2(x_1^0,\cdots,x_n^0)}} ∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx∣∣≤i=1∑mj=1∑nfij2(x10,⋯,xn0)因此, ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ ∣ ∣ ||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||)}{||\Delta x||}|| ∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)∣∣局部有界,因此 lim ∣ ∣ Δ x ∣ ∣ → 0 ∣ ∣ f ′ ( x 1 0 , ⋯ , x n 0 ) Δ x + o 2 ( ∣ ∣ Δ x ∣ ∣ ) ∣ ∣ Δ x ∣ ∣ ∣ ∣ = 0 \lim_{||\Delta x||\to 0}{||\frac{f^\prime(x_1^0,\cdots,x_n^0)\Delta x + o_2(||\Delta x||)}{||\Delta x||}||}=0 ∣∣Δx∣∣→0lim∣∣∣∣Δx∣∣f′(x10,⋯,xn0)Δx+o2(∣∣Δx∣∣)∣∣=0再将(2)代入(1)就可以证得结论
考虑向量函数和多元函数复合的情形:
g
(
y
1
,
⋯
,
y
m
)
g(y_1,\cdots,y_m)
g(y1,⋯,ym)是
m
m
m元函数,
f
(
x
1
,
⋯
,
x
n
)
f(x_1,\cdots,x_n)
f(x1,⋯,xn)是
n
n
n元
m
m
m维向量函数,
g
g
g在
f
(
x
1
0
,
⋯
,
x
n
0
)
f(x_1^0,\cdots,x_n^0)
f(x10,⋯,xn0)处可微,
f
f
f在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微,那么
g
(
f
(
x
1
,
⋯
,
x
n
)
)
g(f(x_1,\cdots,x_n))
g(f(x1,⋯,xn))在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)处可微。
我们记
h
=
g
(
f
)
h=g(f)
h=g(f),设
f
=
(
f
1
,
⋯
,
f
m
)
f=(f_1,\cdots,f_m)
f=(f1,⋯,fm),应用复合函数求导法则,就有:
∂
h
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
i
=
∑
j
=
1
m
∂
f
j
(
x
1
0
,
⋯
,
x
n
0
)
∂
x
i
∂
g
(
y
1
0
,
⋯
,
y
m
0
)
∂
y
j
\frac{\partial h(x_1^0,\cdots,x_n^0)}{\partial x_i} =\sum_{j=1}^m{\frac{\partial f_j(x_1^0,\cdots,x_n^0)}{\partial x_i} \frac{\partial g(y_1^0,\cdots,y_m^0)}{\partial y_j}}
∂xi∂h(x10,⋯,xn0)=j=1∑m∂xi∂fj(x10,⋯,xn0)∂yj∂g(y10,⋯,ym0)这称为多元函数求导的链式法则
高阶偏导数与高阶全微分
高阶偏导就是偏导的偏导,只不过,在高维情形下,由求偏导次序可否交换的问题。以二元函数为例,
f
(
x
,
y
)
f(x,y)
f(x,y)的二阶偏导有四个:
∂
2
f
∂
x
2
\frac{\partial^2 f}{\partial x^2}
∂x2∂2f表示对
x
x
x求两次偏导,
∂
2
f
∂
x
∂
y
\frac{\partial^2 f}{\partial x\partial y}
∂x∂y∂2f表示先对
x
x
x求偏导,再对
y
y
y求偏导,其他两个也可以类似写出。
问题在于
∂
2
f
∂
x
∂
y
=
∂
2
f
∂
y
∂
x
\frac{\partial^2 f}{\partial x\partial y}=\frac{\partial^2 f}{\partial y\partial x}
∂x∂y∂2f=∂y∂x∂2f是否成立?下面我们证明:在高阶偏导数连续的条件下,偏导次序是可以交换的。
定理14.6 f ( x , y ) f(x,y) f(x,y)在 ( x 0 , y 0 ) (x_0,y_0) (x0,y0)的某个邻域上可求二阶偏导数,并且 ∂ 2 f ∂ x ∂ y , ∂ 2 f ∂ y ∂ x \frac{\partial^2 f}{\partial x\partial y},\frac{\partial^2 f}{\partial y\partial x} ∂x∂y∂2f,∂y∂x∂2f都在 ( x 0 , y 0 ) (x_0,y_0) (x0,y0)处连续,则 ∂ 2 f ( x 0 , y 0 ) ∂ x ∂ y = ∂ 2 f ( x 0 , y 0 ) ∂ y ∂ x \frac{\partial^2 f(x_0,y_0)}{\partial x\partial y}= \frac{\partial^2 f(x_0,y_0)}{\partial y\partial x} ∂x∂y∂2f(x0,y0)=∂y∂x∂2f(x0,y0)
证:
首先 [ f ( x , y ) − f ( x , y 0 ) ] − [ f ( x 0 , y ) − f ( x 0 , y 0 ) ] ( x − x 0 ) ( y − y 0 ) = [ f ( x , y ) − f ( x 0 , y ) ] − [ f ( x , y 0 ) − f ( x 0 , y 0 ) ] ( x − x 0 ) ( y − y 0 ) \frac{[f(x,y)-f(x,y_0)]-[f(x_0,y)-f(x_0,y_0)]}{(x-x_0)(y-y_0)} = \frac{[f(x,y)-f(x_0,y)]-[f(x,y_0)-f(x_0,y_0)]}{(x-x_0)(y-y_0)} (x−x0)(y−y0)[f(x,y)−f(x,y0)]−[f(x0,y)−f(x0,y0)]=(x−x0)(y−y0)[f(x,y)−f(x0,y)]−[f(x,y0)−f(x0,y0)]由拉格朗日中值定理,存在 ( 0 , 1 ) (0,1) (0,1)之间的正实数 θ 1 \theta_1 θ1和 θ 2 \theta_2 θ2 [ f ( x , y ) − f ( x , y 0 ) ] − [ f ( x 0 , y ) − f ( x 0 , y 0 ) ] ( x − x 0 ) ( y − y 0 ) = f x ( x 0 + θ 1 ( x − x 0 ) , y ) − f x ( x 0 + θ 1 ( x − x 0 ) , y 0 ) y − y 0 = f x y ( x 0 + θ 1 ( x − x 0 ) , y 0 + θ 2 ( y − y 0 ) ) \frac{[f(x,y)-f(x,y_0)]-[f(x_0,y)-f(x_0,y_0)]}{(x-x_0)(y-y_0)} \\= \frac{ f_x(x_0+\theta_1(x-x_0),y)-f_x(x_0+\theta_1(x-x_0),y_0) }{y-y_0}\\=f_{xy}(x_0+\theta_1(x-x_0),y_0+\theta_2(y-y_0)) (x−x0)(y−y0)[f(x,y)−f(x,y0)]−[f(x0,y)−f(x0,y0)]=y−y0fx(x0+θ1(x−x0),y)−fx(x0+θ1(x−x0),y0)=fxy(x0+θ1(x−x0),y0+θ2(y−y0)) h ( x , y ) = [ f ( x , y ) − f ( x , y 0 ) ] − [ f ( x 0 , y ) − f ( x 0 , y 0 ) ] ( x − x 0 ) ( y − y 0 ) h(x,y)=\frac{[f(x,y)-f(x,y_0)]-[f(x_0,y)-f(x_0,y_0)]}{(x-x_0)(y-y_0)} h(x,y)=(x−x0)(y−y0)[f(x,y)−f(x,y0)]−[f(x0,y)−f(x0,y0)] h ( x , y ) h(x,y) h(x,y)的全面极限和两个累次极限存在,相等,这样就可以证得偏导次序可交换
我们假设二元函数在
(
x
0
,
y
0
)
(x_0,y_0)
(x0,y0)的某个邻域上各阶偏导数都存在,那么,各阶偏导数都在
(
x
0
,
y
0
)
(x_0,y_0)
(x0,y0)处连续(因为连续可微),考察函数:
h
(
t
)
=
f
(
x
0
+
t
,
y
0
+
t
)
h(t)=f(x_0+t,y_0+t)
h(t)=f(x0+t,y0+t),则
h
(
t
)
h(t)
h(t)在
t
=
0
t=0
t=0处的各阶偏导数:
h
(
k
)
(
0
)
=
∑
i
=
0
k
C
k
i
∂
k
f
(
x
0
,
y
0
)
∂
x
i
∂
y
(
k
−
i
)
h^{(k)}(0)=\sum_{i=0}^{k}{C_k^i \frac{\partial^k f(x_0,y_0)}{\partial x^i \partial y^{(k-i)}}}
h(k)(0)=i=0∑kCki∂xi∂y(k−i)∂kf(x0,y0)这就和二项式定理类似,其他高阶导数的求法,大多用到数学归纳法,这里不再赘述。
方向导数
在一元情况下,导数有左导数和右导数,而在多元情形下,由于方向远远不止两个,但我们还是可以定义出方向导数。
方向向量就定义为
d
=
(
d
1
,
⋯
,
d
n
)
d=(d_1,\cdots,d_n)
d=(d1,⋯,dn),其中
∑
k
=
1
n
d
k
2
=
1
\sqrt{\sum_{k=1}^n{d_k^2}}=1
∑k=1ndk2=1,
d
d
d就称为方向向量。方向导数就定义为极限:
∂
f
(
x
)
∂
d
=
lim
t
→
0
+
f
(
x
+
t
d
)
−
f
(
x
)
t
\frac{\partial f(x)}{\partial d}=\lim_{t\to 0^+}{\frac{f(x+td)-f(x)}{t}}
∂d∂f(x)=t→0+limtf(x+td)−f(x)方向导数该如何计算呢?如果
f
(
x
)
f(x)
f(x)在
f
(
x
0
)
f(x_0)
f(x0)处可微,那么
f
(
x
0
+
t
d
)
−
f
(
x
0
)
=
t
f
′
(
x
0
)
d
T
+
o
(
t
)
f(x_0+td)-f(x_0)=tf^\prime(x_0)d^T+o(t)
f(x0+td)−f(x0)=tf′(x0)dT+o(t)这样
∂
f
(
x
0
)
∂
d
=
f
′
(
x
0
)
d
T
\frac{\partial f(x_0)}{\partial d} = f^\prime(x_0)d^T
∂d∂f(x0)=f′(x0)dT实际上就是偏导按照方向进行加权。
高维泰勒公式
高维泰勒公式,就是
f
(
x
0
+
t
(
x
−
x
0
)
)
f(x_0+t(x-x_0))
f(x0+t(x−x0))在
0
0
0处的泰勒公式,再令
t
=
1
t=1
t=1,高维泰勒公式形式比较复杂,在三阶以上很难写出一般的形式。不过,我们这里给出零阶,一阶,二阶的泰勒公式的形式,在多元函数极值判断中起到重要的作用。我们称矩阵
H
f
(
x
0
)
=
[
f
11
(
x
0
)
f
12
(
x
0
)
⋯
f
1
n
(
x
0
)
f
21
(
x
0
)
f
22
(
x
0
)
⋯
f
2
n
(
x
0
)
⋯
⋯
⋯
⋯
f
n
1
(
x
0
)
f
n
2
(
x
0
)
⋯
f
n
n
(
x
0
)
]
H_f(x_0) = \left[ \begin{matrix} f_{11}(x_0)&f_{12}(x_0)&\cdots&f_{1n}(x_0)\\ f_{21}(x_0)&f_{22}(x_0)&\cdots&f_{2n}(x_0)\\ \cdots&\cdots&\cdots&\cdots\\ f_{n1}(x_0)&f_{n2}(x_0)&\cdots&f_{nn}(x_0) \end{matrix}\right]
Hf(x0)=⎣⎢⎢⎡f11(x0)f21(x0)⋯fn1(x0)f12(x0)f22(x0)⋯fn2(x0)⋯⋯⋯⋯f1n(x0)f2n(x0)⋯fnn(x0)⎦⎥⎥⎤为
f
f
f的海色矩阵,海色矩阵在多元函数情形下,起到二阶导数的作用。令
x
=
(
x
1
,
⋯
,
x
n
)
,
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
x=(x_1,\cdots,x_n),x_0=(x_1^0,\cdots,x_n^0)
x=(x1,⋯,xn),x0=(x10,⋯,xn0),令
g
(
t
)
=
f
(
x
0
+
t
(
x
−
x
0
)
)
g(t)=f(x_0+t(x-x_0))
g(t)=f(x0+t(x−x0)),由复合函数求导法则,则
g
′
(
t
)
=
∑
i
=
1
n
(
x
i
−
x
i
0
)
f
i
′
(
x
0
+
t
(
x
−
x
0
)
)
g
′
′
(
t
)
=
∑
i
=
1
n
∑
j
=
1
n
(
x
i
−
x
i
0
)
(
x
j
−
x
j
0
)
f
i
j
′
′
(
x
0
+
t
(
x
−
x
0
)
)
g^\prime(t)=\sum_{i=1}^n(x_i-x_i^0)f_i^\prime(x_0+t(x-x_0))\\ g^{\prime\prime}(t)=\sum_{i=1}^n\sum_{j=1}^n(x_i-x_i^0)(x_j-x_j^0)f_{ij}^{\prime\prime}(x_0+t(x-x_0))
g′(t)=i=1∑n(xi−xi0)fi′(x0+t(x−x0))g′′(t)=i=1∑nj=1∑n(xi−xi0)(xj−xj0)fij′′(x0+t(x−x0))则
g
(
t
)
g(t)
g(t)的零阶麦克劳林公式为
g
(
t
)
−
g
(
0
)
=
∑
i
=
1
n
(
x
i
−
x
i
0
)
f
i
′
(
x
0
+
ξ
(
x
−
x
0
)
)
t
g(t)-g(0)=\sum_{i=1}^n(x_i-x_i^0)f_i^\prime(x_0+\xi(x-x_0))t
g(t)−g(0)=i=1∑n(xi−xi0)fi′(x0+ξ(x−x0))t其中
ξ
∈
(
0
,
t
)
\xi\in(0,t)
ξ∈(0,t),令
t
=
1
t=1
t=1,有
f
(
x
)
−
f
(
x
0
)
=
∑
i
=
1
n
f
i
′
(
x
0
+
ξ
(
x
−
x
0
)
)
(
x
i
−
x
i
0
)
f(x)-f(x_0)=\sum_{i=1}^nf_i^\prime(x_0+\xi(x-x_0))(x_i-x_i^0)
f(x)−f(x0)=i=1∑nfi′(x0+ξ(x−x0))(xi−xi0)其中
ξ
∈
(
0
,
1
)
\xi\in(0,1)
ξ∈(0,1),可以看出
x
′
=
x
0
+
ξ
(
x
−
x
0
)
x^\prime=x_0+\xi(x-x_0)
x′=x0+ξ(x−x0)是连接
x
0
x_0
x0和
x
x
x的线段上的一点,也就是说,如果
f
f
f连续可微,则对任意的
x
1
,
x
2
x_1,x_2
x1,x2,存在连接
x
1
,
x
2
x_1,x_2
x1,x2连段上的一点
x
3
x_3
x3,使得
f
(
x
1
)
−
f
(
x
2
)
=
∑
i
=
1
n
f
i
′
(
x
3
)
(
x
i
1
−
x
i
2
)
f(x_1)-f(x_2)=\sum_{i=1}^nf_i^\prime(x_3)(x_i^1-x_i^2)
f(x1)−f(x2)=i=1∑nfi′(x3)(xi1−xi2)这就是
n
n
n维空间上的拉格朗日中值定理,同样可以写出一阶的泰勒公式
f
(
x
)
−
f
(
x
0
)
=
∑
i
=
1
n
f
i
′
(
x
0
)
(
x
i
−
x
i
0
)
+
1
2
(
x
−
x
0
)
T
H
f
(
x
0
+
ξ
(
x
−
x
0
)
)
(
x
−
x
0
)
f(x)-f(x_0)=\sum_{i=1}^nf_i^\prime(x_0)(x_i-x_i^0)+\frac{1}{2}(x-x_0)^TH_f(x_0+\xi(x-x_0))(x-x_0)
f(x)−f(x0)=i=1∑nfi′(x0)(xi−xi0)+21(x−x0)THf(x0+ξ(x−x0))(x−x0)二阶泰勒公式为
f
(
x
)
=
f
(
x
0
)
+
f
′
(
x
0
)
(
x
−
x
0
)
+
1
2
(
x
−
x
0
)
T
H
f
(
x
0
)
(
x
−
x
0
)
+
o
(
(
x
−
x
0
)
T
(
x
−
x
0
)
)
f(x) = f(x_0)+f^\prime(x_0)(x-x_0)+\frac{1}{2}(x-x_0)^T H_f(x_0)(x-x_0) + o((x-x_0)^T(x-x_0))
f(x)=f(x0)+f′(x0)(x−x0)+21(x−x0)THf(x0)(x−x0)+o((x−x0)T(x−x0))
隐函数存在定理
单个方程情形
现在,我们来讨论多元微分学最重要的定理之一,隐函数存在定理。实际上,函数,可由显方程给出,即:
y
=
f
(
x
1
,
⋯
,
x
n
)
y=f(x_1,\cdots,x_n)
y=f(x1,⋯,xn)然而,更多情况下,函数并没有以上明确的表达式,而是通常以方程的形式给出
F
(
y
,
x
1
,
⋯
,
x
n
)
=
0
F(y,x_1,\cdots,x_n)=0
F(y,x1,⋯,xn)=0这种形式给出的函数,我们称为隐函数,并且,很多情况下,我们根本无法从隐函数方程反解出一个函数出来,那么,我们怎么确认是否存在一个函数,满足以上方程呢?隐函数存在定理,就是要回答上面的问题。首先我们需要理清概念,何谓隐含函数:对
n
+
1
n+1
n+1元函数
F
(
x
1
,
⋯
,
x
n
,
y
)
F(x_1,\cdots,x_n,y)
F(x1,⋯,xn,y),方程
F
(
x
1
,
⋯
,
x
n
,
y
)
=
0
F(x_1,\cdots,x_n,y)=0
F(x1,⋯,xn,y)=0确定一个隐函数
y
=
f
(
x
1
,
⋯
,
x
n
)
y=f(x_1,\cdots,x_n)
y=f(x1,⋯,xn)的含义的是对定义域
E
⊂
R
n
E\subset R^n
E⊂Rn上的点
x
∈
E
x\in E
x∈E,都有
F
(
x
,
f
(
x
)
)
=
0
F(x,f(x))=0
F(x,f(x))=0首先要弄清楚的问题是函数
f
f
f是否存在,其次是否唯一。举一例说明:
x
2
+
y
2
=
1
x^2+y^2=1
x2+y2=1在定义域
[
−
1
,
1
]
[-1,1]
[−1,1]上确实存在隐函数
y
=
1
−
x
2
y=\sqrt{1-x^2}
y=1−x2和
y
=
−
1
−
x
2
y=-\sqrt{1-x^2}
y=−1−x2可见,隐函数虽然存在,但是不唯一,两个隐函数的区别在于其值域。可见,在隐函数的问题上,不仅要指定在哪个定义域范围内存在隐函数,
还
要
确
定
因
变
量
的
范
围
‾
\underline{还要确定因变量的范围}
还要确定因变量的范围,否则隐函数可能不唯一。
我们先假设隐函数存在,并且,
F
F
F对各个变元可求连续偏导,并且隐函数也对各个变元可求连续的偏导数,我们就想到,直接对方程两边求导:
∂
y
∂
x
i
F
y
′
+
F
i
′
=
0
\frac{\partial y}{\partial x_i}F_y ^\prime+ F_{i}^\prime = 0
∂xi∂yFy′+Fi′=0只要
F
y
′
≠
0
F_y^\prime\neq 0
Fy′=0,就可以求出
∂
y
∂
x
i
\frac{\partial y}{\partial x_i}
∂xi∂y。隐函数存在定理将表明,只要
F
y
′
≠
0
F_y^\prime\neq 0
Fy′=0,在很小的一个局部,隐函数是存在的,并且确定的隐函数还是连续可微的。接下来我们用的范数是
∣
(
x
1
,
x
2
,
⋯
,
x
n
)
∣
=
max
1
≤
i
≤
n
∣
x
i
∣
\displaystyle|(x_1,x_2,\cdots,x_n)|=\max_{1\le i\le n}|x_i|
∣(x1,x2,⋯,xn)∣=1≤i≤nmax∣xi∣,容易验证这是一个范数,这个范数的好处是,由其诱导的距离产生的邻域是一个方体,即对任意的
x
0
=
(
x
1
(
0
)
,
⋯
,
x
n
(
0
)
)
x_0=(x_1^{(0)},\cdots,x_n^{(0)})
x0=(x1(0),⋯,xn(0)),则
B
(
x
0
,
δ
)
=
{
(
x
1
,
⋯
,
x
n
)
:
∣
x
i
−
x
i
(
0
)
∣
<
δ
,
i
=
1
,
⋯
,
n
}
B(x_0,\delta)=\{(x_1,\cdots,x_n):|x_i-x_i^{(0)}|<\delta,i=1,\cdots,n\}
B(x0,δ)={(x1,⋯,xn):∣xi−xi(0)∣<δ,i=1,⋯,n}由此,如果我们将
x
0
=
(
x
0
′
,
x
0
′
′
)
x_0=(x_0^\prime,x_0^{\prime\prime})
x0=(x0′,x0′′),其中
x
0
′
=
(
x
1
(
0
)
,
⋯
,
x
m
(
0
)
)
x_0^\prime=(x_1^{(0)},\cdots,x_m^{(0)})
x0′=(x1(0),⋯,xm(0))。则
B
(
x
0
,
δ
)
=
B
(
x
0
′
,
δ
)
×
B
(
x
0
′
′
,
δ
)
B(x_0,\delta)=B(x_0^\prime,\delta)\times B(x_0^{\prime\prime},\delta)
B(x0,δ)=B(x0′,δ)×B(x0′′,δ)这比较方便我们书写证明的过程
定理14.7(隐函数存在定理1)
F
(
x
1
,
⋯
,
x
n
,
y
)
F(x_1,\cdots,x_n,y)
F(x1,⋯,xn,y)是某个开集
G
G
G上的连续可微函数,并且
(
x
1
0
,
⋯
,
x
n
0
,
y
0
)
∈
G
(x_1^0,\cdots,x_n^0,y^0)\in G
(x10,⋯,xn0,y0)∈G,同时满足:
(1)
F
(
x
1
0
,
⋯
,
x
n
0
,
y
0
)
=
0
F(x_1^0,\cdots,x_n^0,y_0)=0
F(x10,⋯,xn0,y0)=0
(2)
F
y
(
x
1
0
,
⋯
,
x
n
0
,
y
0
)
≠
0
F_y(x_1^0,\cdots,x_n^0,y_0)\neq 0
Fy(x10,⋯,xn0,y0)=0
则存在
(
x
1
0
,
⋯
,
x
n
0
)
(x_1^0,\cdots,x_n^0)
(x10,⋯,xn0)的邻域
B
1
B_1
B1,
y
0
y_0
y0的邻域
B
B
B,
B
1
×
B
⊂
G
B_1\times B \subset G
B1×B⊂G,存在定义在
B
1
B_1
B1上的唯一的
n
n
n元连续可微函数
y
=
f
(
x
1
,
⋯
,
x
n
)
y=f(x_1,\cdots,x_n)
y=f(x1,⋯,xn),满足:
(1)
f
(
x
1
,
⋯
,
x
n
)
∈
B
∀
(
x
1
,
⋯
,
x
n
)
∈
B
1
f(x_1,\cdots,x_n)\in B \quad \forall (x_1,\cdots,x_n) \in B_1
f(x1,⋯,xn)∈B∀(x1,⋯,xn)∈B1
(2)
y
0
=
f
(
x
1
0
,
⋯
,
x
n
0
)
y_0=f(x_1^0,\cdots,x_n^0)
y0=f(x10,⋯,xn0)
(3)
F
(
x
1
,
⋯
,
x
n
,
f
(
x
1
,
⋯
,
x
n
)
)
=
0
∀
(
x
1
,
⋯
,
x
n
)
∈
B
1
F(x_1,\cdots,x_n,f(x_1,\cdots,x_n))=0\quad \forall (x_1,\cdots,x_n) \in B_1
F(x1,⋯,xn,f(x1,⋯,xn))=0∀(x1,⋯,xn)∈B1
(4)
y
y
y在
B
1
B_1
B1上连续可微,并且
∂
y
∂
x
k
=
−
F
x
k
′
F
y
′
\frac{\partial y}{\partial x_k} = -\frac{F_{x_k}^\prime}{F_y^\prime}
∂xk∂y=−Fy′Fxk′
证:
这里假设 F y ′ ( x 1 0 , ⋯ , x n 0 , y 0 ) > 0 F_y^\prime(x_1^0,\cdots,x_n^0,y_0)>0 Fy′(x10,⋯,xn0,y0)>0,而 F y ′ ( x 1 0 , ⋯ , x n 0 , y 0 ) < 0 F_y^\prime(x_1^0,\cdots,x_n^0,y_0)<0 Fy′(x10,⋯,xn0,y0)<0的证明是类似的。
第 一 步 ‾ \underline{第一步} 第一步:我们要证明隐函数的存在性:
由偏导的连续性,存在 δ > 0 \delta >0 δ>0,当 ( x , y ) ∈ B ( ( x 0 , y 0 ) , δ ) ⊂ G (x,y)\in B((x_0,y_0),\delta) \subset G (x,y)∈B((x0,y0),δ)⊂G时
F y ′ ( x 1 , ⋯ , x n , y ) > 0 F_y^\prime(x_1,\cdots,x_n,y)>0 Fy′(x1,⋯,xn,y)>0当 ( x , y ) ∈ B ( ( x 0 , y 0 ) , δ ) (x,y) \in B((x_0,y_0),\delta) (x,y)∈B((x0,y0),δ)时, F ( x 1 , ⋯ , x n , y ) F(x_1,\cdots,x_n,y) F(x1,⋯,xn,y)作为 y y y的一元函数在 ( y 0 − δ , y 0 + δ ) (y_0-\delta,y_0+\delta) (y0−δ,y0+δ)上严格单调上升。
于是,有 { F ( x 1 0 , ⋯ , x n 0 , y 0 + δ 2 ) > 0 F ( x 1 0 , ⋯ , x n 0 , y 0 − δ 2 ) < 0 \begin{cases} F(x_1^0,\cdots,x_n^0,y_0+\frac{\delta}{2})>0\\ F(x_1^0,\cdots,x_n^0,y_0-\frac{\delta}{2})<0 \end{cases} {F(x10,⋯,xn0,y0+2δ)>0F(x10,⋯,xn0,y0−2δ)<0由函数的连续性,存在 0 < δ 2 < δ 2 0<\delta_2<\frac{\delta}{2} 0<δ2<2δ, ( x , y ) ∈ B ( ( x 0 , y 0 + δ 2 ) , δ 2 ) (x,y)\in B((x_0,y_0+\frac{\delta}{2}),\delta_2) (x,y)∈B((x0,y0+2δ),δ2)时,就有
F ( x 1 , ⋯ , x n , y ) > 0 F(x_1,\cdots,x_n,y)>0 F(x1,⋯,xn,y)>0并且对任意的 ( x , y ) ∈ B ( ( x 0 , y 0 + δ 2 ) , δ 2 ) (x,y)\in B((x_0,y_0+\frac{\delta}{2}),\delta_2) (x,y)∈B((x0,y0+2δ),δ2),就有 d ( ( x , y ) , ( x 0 , y 0 ) ) ≤ d ( ( x , y ) , ( x 0 , y 0 + δ 2 ) ) + d ( ( x 0 , y 0 + δ 2 ) , ( x 0 , y 0 ) ) < δ 2 + δ 2 < δ \begin{aligned} &d((x,y),(x_0,y_0))\\ \le& d((x,y),(x_0,y_0+\frac{\delta}{2}))+d((x_0,y_0+\frac{\delta}{2}),(x_0,y_0))<\delta_2+\frac{\delta}{2}<\delta \end{aligned} ≤d((x,y),(x0,y0))d((x,y),(x0,y0+2δ))+d((x0,y0+2δ),(x0,y0))<δ2+2δ<δ即 B ( ( x 0 , y 0 + δ 2 ) , δ 2 ) ⊂ B ( ( x 0 , y 0 ) , δ ) B((x_0,y_0+\frac{\delta}{2}),\delta_2)\subset B((x_0,y_0),\delta) B((x0,y0+2δ),δ2)⊂B((x0,y0),δ),只要 x ∈ B ( ( x 0 , y 0 − δ 2 ) , δ 2 ) x\in B((x_0,y_0-\frac{\delta}{2}),\delta_2) x∈B((x0,y0−2δ),δ2),就有 F ( x 1 , ⋯ , x n , y ) < 0 F(x_1,\cdots,x_n,y)<0 F(x1,⋯,xn,y)<0同理, B ( ( x 0 , y 0 − δ 2 ) , δ 2 ) ⊂ B ( ( x 0 , y 0 ) , δ ) B((x_0,y_0-\frac{\delta}{2}),\delta_2)\subset B((x_0,y_0),\delta) B((x0,y0−2δ),δ2)⊂B((x0,y0),δ)。只要 x ∈ B ( x 0 , δ 2 ) x\in B(x_0,\delta_2) x∈B(x0,δ2),就有 { F ( x 1 , ⋯ , x n , y 0 + δ 2 ) > 0 F ( x 1 , ⋯ , x n , y 0 − δ 2 ) < 0 \begin{cases} F(x_1,\cdots,x_n,y_0+\frac{\delta}{2})>0\\ F(x_1,\cdots,x_n,y_0-\frac{\delta}{2})<0 \end{cases} {F(x1,⋯,xn,y0+2δ)>0F(x1,⋯,xn,y0−2δ)<0并且,当 y 0 − δ 2 ≤ y ≤ y 0 + δ 2 y_0-\frac{\delta}{2}\le y \le y_0+\frac{\delta}{2} y0−2δ≤y≤y0+2δ时
d ( ( x , y ) , ( x 0 , y 0 ) ) < δ d((x,y),(x_0,y_0))<\delta d((x,y),(x0,y0))<δ F ( x 1 , ⋯ , x n , y ) F(x_1,\cdots,x_n,y) F(x1,⋯,xn,y)作为 y y y的一元函数,在 [ y 0 − δ 2 , y 0 + δ 2 ] [y_0-\frac{\delta}{2},y_0+\frac{\delta}{2}] [y0−2δ,y0+2δ]上连续,由介值定理,存在 f ( x 1 , ⋯ , x n ) ∈ ( y 0 − δ 2 , y 0 + δ 2 ) f(x_1,\cdots,x_n)\in (y_0-\frac{\delta}{2},y_0+\frac{\delta}{2}) f(x1,⋯,xn)∈(y0−2δ,y0+2δ)
F ( x 1 , ⋯ , x n , f ( x 1 , ⋯ , x n ) ) = 0 F(x_1,\cdots,x_n,f(x_1,\cdots,x_n))=0 F(x1,⋯,xn,f(x1,⋯,xn))=0并且 ( x 1 , x 2 , ⋯ , x n , f ( x 1 , ⋯ , x n ) ) ∈ B ( ( x 0 , y 0 ) , δ ) (x_1,x_2,\cdots,x_n,f(x_1,\cdots,x_n))\in B((x_0,y_0),\delta) (x1,x2,⋯,xn,f(x1,⋯,xn))∈B((x0,y0),δ),这就证得了隐函数的存在性,令 B = B ( y 0 , δ ) B=B(y_0,\delta) B=B(y0,δ), B 1 = B ( x 0 , δ 2 ) B_1=B(x_0,\delta_2) B1=B(x0,δ2),则 f f f满足:
(1) f ( x 0 ) = y 0 f(x_0)=y_0 f(x0)=y0
(2) F ( x , f ( x ) ) = 0 , f ( x ) ∈ B , ∀ x ∈ B 1 F(x,f(x))=0,f(x)\in B,\forall x \in B_1 F(x,f(x))=0,f(x)∈B,∀x∈B1
并且这个函数 f f f是唯一的,假设存在定义在 B 1 = B ( x 0 , δ 2 ) B_1=B(x_0,\delta_2) B1=B(x0,δ2)上的函数 g g g,也满足:
(1) g ( x 0 ) = y 0 g(x_0)=y_0 g(x0)=y0
(2) F ( x , g ( x ) ) = 0 , g ( x ) ∈ B , ∀ x ∈ B 1 F(x,g(x))=0,g(x)\in B,\forall x \in B_1 F(x,g(x))=0,g(x)∈B,∀x∈B1
则对任意的 x ∈ B 1 x\in B_1 x∈B1,那么由于 F ( x , y ) F(x,y) F(x,y)作为 y y y的一元函数在 ( y 0 − δ , y 0 + δ ) (y_0-\delta,y_0+\delta) (y0−δ,y0+δ)上严格单调上升,如果 g ( x ) > f ( x ) g(x) > f(x) g(x)>f(x),那么, F ( x , f ( x ) ) = 0 < F ( x , g ( x ) ) F(x,f(x))=0<F(x,g(x)) F(x,f(x))=0<F(x,g(x)),矛盾,因此 g ( x ) ≤ f ( x ) g(x)\le f(x) g(x)≤f(x),同理可证 g ( x ) ≥ f ( x ) g(x) \ge f(x) g(x)≥f(x),因此 f ( x ) = g ( x ) f(x)=g(x) f(x)=g(x),这证明了隐函数的唯一性,同时 B 1 × B ⊂ B ( ( x 0 , y 0 ) , δ ) ⊂ G B_1\times B \subset B((x_0,y_0),\delta)\subset G B1×B⊂B((x0,y0),δ)⊂G。
第 二 步 ‾ \underline{第二步} 第二步:证明 f f f在 B 1 B_1 B1上是连续的。对任意的 x ∈ B 2 x\in B_2 x∈B2,令 ε 0 = δ − ∣ f ( x ) − y 0 ∣ \varepsilon_0=\delta-|f(x)-y_0| ε0=δ−∣f(x)−y0∣,对任意的 ε > 0 \varepsilon>0 ε>0,不妨设 ε < ε 0 \varepsilon<\varepsilon_0 ε<ε0,此时, f ( x ) + ε < f ( x ) + δ − ∣ f ( x ) − y 0 ∣ ≤ y 0 + δ f(x)+\varepsilon<f(x)+\delta-|f(x)-y_0|\le y_0+\delta f(x)+ε<f(x)+δ−∣f(x)−y0∣≤y0+δ, f ( x ) − ε > f ( x ) − δ + ∣ f ( x ) − y 0 ∣ ≥ y 0 − δ f(x)-\varepsilon>f(x)-\delta+|f(x)-y_0|\ge y_0-\delta f(x)−ε>f(x)−δ+∣f(x)−y0∣≥y0−δ。由于 F ( x , y ) F(x,y) F(x,y)作为 y y y的一元函数在 ( f ( x ) − ε , f ( x ) + ε ) (f(x)-\varepsilon,f(x)+\varepsilon) (f(x)−ε,f(x)+ε)上严格单调上升,从而 { F ( x , f ( x ) − ε ) < 0 F ( x , f ( x ) + ε ) > 0 \begin{cases} F(x,f(x)-\varepsilon)<0\\ F(x,f(x)+\varepsilon)>0 \end{cases} {F(x,f(x)−ε)<0F(x,f(x)+ε)>0由 F F F的连续性, 存在 δ 3 > 0 \delta_3>0 δ3>0, δ 3 < min ( δ − d ( ( x 0 , y 0 ) , ( x , f ( x ) − ε ) ) , δ − d ( ( x 0 , y 0 ) , ( x , f ( x ) + ε ) ) , δ 2 − d ( x , x 0 ) ) \delta_3<\min(\delta-d((x_0,y_0),(x,f(x)-\varepsilon)),\delta-d((x_0,y_0),(x,f(x)+\varepsilon)),\delta_2-d(x,x_0)) δ3<min(δ−d((x0,y0),(x,f(x)−ε)),δ−d((x0,y0),(x,f(x)+ε)),δ2−d(x,x0)),当 ( x ′ , y ′ ) ∈ B ( ( x , f ( x ) − ε ) , δ 3 ) (x^\prime,y^\prime)\in B((x,f(x)-\varepsilon),\delta_3) (x′,y′)∈B((x,f(x)−ε),δ3)时 d ( x ′ , x 0 ) ≤ d ( x ′ , x ) + d ( x , x 0 ) < δ 2 ∣ y ′ − y 0 ∣ ≤ ∣ y ′ − f ( x ) + ε ∣ + ∣ f ( x ) − ε − y 0 ∣ < δ d(x^\prime,x_0)\le d(x^\prime,x)+d(x,x_0)<\delta_2\\ |y^\prime-y_0|\le |y^\prime-f(x)+\varepsilon|+|f(x)-\varepsilon-y_0|<\delta d(x′,x0)≤d(x′,x)+d(x,x0)<δ2∣y′−y0∣≤∣y′−f(x)+ε∣+∣f(x)−ε−y0∣<δ有 F ( x , y ) < 0 F(x,y)<0 F(x,y)<0当 ( x ′ , y ′ ) ∈ B ( ( x , f ( x ) + ε ) , δ 3 ) (x^\prime,y^\prime)\in B((x,f(x)+\varepsilon),\delta_3) (x′,y′)∈B((x,f(x)+ε),δ3)时 d ( x ′ , x 0 ) ≤ d ( x ′ , x ) + d ( x , x 0 ) < δ 2 ∣ y ′ − y 0 ∣ ≤ ∣ y ′ − f ( x ) − ε ∣ + ∣ f ( x ) + ε − y 0 ∣ < δ d(x^\prime,x_0)\le d(x^\prime,x)+d(x,x_0)<\delta_2\\ |y^\prime-y_0|\le |y^\prime-f(x)-\varepsilon|+|f(x)+\varepsilon-y_0|<\delta d(x′,x0)≤d(x′,x)+d(x,x0)<δ2∣y′−y0∣≤∣y′−f(x)−ε∣+∣f(x)+ε−y0∣<δ有 F ( x , y ) > 0 F(x,y)>0 F(x,y)>0则当 d ( x ′ , x ) < δ 3 d(x^\prime,x)<\delta_3 d(x′,x)<δ3时,有 F ( x ′ , f ( x ) + ε ) > 0 F ( x ′ , f ( x ) − ε ) < 0 F(x^\prime,f(x)+\varepsilon)>0\\ F(x^\prime,f(x)-\varepsilon)<0 F(x′,f(x)+ε)>0F(x′,f(x)−ε)<0从而由介值定理, 存在 y ′ ∈ ( f ( x ) − ε , f ( x ) + ε ) y^\prime\in (f(x)-\varepsilon,f(x)+\varepsilon) y′∈(f(x)−ε,f(x)+ε),满足 F ( x ′ , y ′ ) = 0 F(x^\prime,y^\prime)=0 F(x′,y′)=0再由隐函数的唯一性, y ′ = f ( x ′ ) y^\prime=f(x^\prime) y′=f(x′),从而 ∣ f ( x ′ ) − f ( x ) ∣ < ε |f(x^\prime)-f(x)|<\varepsilon ∣f(x′)−f(x)∣<ε 第 三 步 ‾ : \underline{第三步}: 第三步:证明 y = f ( x ) y=f(x) y=f(x)是连续可微的。
对任意的 x ∈ B 1 x\in B_1 x∈B1,对 k = 1 , ⋯ , n k=1,\cdots,n k=1,⋯,n,对任意的 0 < Δ x k < δ 2 − d ( x , x 0 ) 0<\Delta x_k<\delta_2-d(x,x_0) 0<Δxk<δ2−d(x,x0),则令 x ′ = ( x 1 , ⋯ , x k − 1 , x k + Δ x k , x k + 1 , x n ) x^\prime = (x_1,\cdots,x_{k-1},x_k+\Delta x_k,x_{k+1},x_n) x′=(x1,⋯,xk−1,xk+Δxk,xk+1,xn),则 d ( x ′ , x 0 ) ≤ d ( x ′ , x ) + d ( x , x 0 ) = Δ x k + d ( x , x 0 ) < δ 2 d(x^\prime,x_0)\le d(x^\prime,x)+d(x,x_0)=\Delta x_k + d(x,x_0)<\delta_2 d(x′,x0)≤d(x′,x)+d(x,x0)=Δxk+d(x,x0)<δ2则存在 ξ ∈ ( 0 , 1 ) \xi\in(0,1) ξ∈(0,1),令 x ′ ′ = ( x 1 , ⋯ , x k − 1 , x k + ξ Δ x k , x k + 1 , ⋯ , x n ) x^{\prime\prime}=(x_1,\cdots,x_{k-1},x_k+\xi\Delta x_k,x_{k+1},\cdots,x_n) x′′=(x1,⋯,xk−1,xk+ξΔxk,xk+1,⋯,xn),使得 F ( x ′ , f ( x ′ ) ) − F ( x , f ( x ) ) = F x k ′ ( x ′ ′ ) Δ x k + F y ′ ( x ′ ′ ) [ f ( x ′ ) − f ( x ) ] = 0 F(x^\prime,f(x^\prime))-F(x,f(x))=F_{x_k}^\prime(x^{\prime\prime})\Delta x_k+F_y^\prime(x^{\prime\prime})[f(x^\prime)-f(x)]=0 F(x′,f(x′))−F(x,f(x))=Fxk′(x′′)Δxk+Fy′(x′′)[f(x′)−f(x)]=0则 f ( x ′ ) − f ( x ) Δ x k = − F x k ′ ( x ′ ′ ) F y ′ ( x ′ ′ ) \frac{f(x^\prime)-f(x)}{\Delta x_k}=-\frac{F_{x_k}^\prime(x^{\prime\prime})}{F_{y}^\prime(x^{\prime\prime})} Δxkf(x′)−f(x)=−Fy′(x′′)Fxk′(x′′)当 Δ x k → 0 \Delta x_k \to 0 Δxk→0时, x ′ ′ → x x^{\prime\prime} \to x x′′→x,从而有 lim Δ x k → 0 f ( x ′ ) − f ( x ) Δ x k = − lim Δ x k → 0 F x k ′ ( x ′ ′ ) F y ′ ( x ′ ′ ) = − F x k ′ ( x ) F y ′ ( x ) \lim_{\Delta x_k\to 0}\frac{f(x^\prime)-f(x)}{\Delta x_k}=-\lim_{\Delta x_k\to 0}\frac{F_{x_k}^\prime(x^{\prime\prime})}{F_{y}^\prime(x^{\prime\prime})}=-\frac{F_{x_k}^\prime(x)}{F_y^\prime(x)} Δxk→0limΔxkf(x′)−f(x)=−Δxk→0limFy′(x′′)Fxk′(x′′)=−Fy′(x)Fxk′(x)因此 f x k ′ ( x ) = − F x k ′ ( x ) F y ′ ( x ) f_{x_k}^\prime(x)=-\frac{F_{x_k}^\prime(x)}{F_y^\prime(x)} fxk′(x)=−Fy′(x)Fxk′(x)由于 F x k ′ , F y ′ F_{x_k}^\prime,F_y^\prime Fxk′,Fy′都是连续的,因此 f x k ′ ( x ) f_{x_k}^\prime(x) fxk′(x)是连续的,从而 f ( x ) f(x) f(x)是连续可微。
也就是说,只要对 y y y的偏导数不为0,那么,就算我们从代数上无法反解出隐函数,也可以确认隐函数一定是存在的,并且还是连续可微的,那么我们上面求隐函数偏导的做法是合理的。
方程组情形
很多情况下,我们面对的是一个隐函数组:
{
F
1
(
x
1
,
⋯
,
x
n
,
y
1
,
⋯
,
y
m
)
=
0
F
2
(
x
1
,
⋯
,
x
n
,
y
1
,
⋯
,
y
m
)
=
0
⋯
F
m
(
x
1
,
⋯
,
x
n
,
y
1
,
⋯
,
y
m
)
=
0
(5)
\tag{5} \begin{cases} F_1(x_1,\cdots,x_n,y_1,\cdots,y_m)=0\\ F_2(x_1,\cdots,x_n,y_1,\cdots,y_m)=0\\ \cdots\\ F_m(x_1,\cdots,x_n,y_1,\cdots,y_m)=0 \end{cases}
⎩⎪⎪⎪⎨⎪⎪⎪⎧F1(x1,⋯,xn,y1,⋯,ym)=0F2(x1,⋯,xn,y1,⋯,ym)=0⋯Fm(x1,⋯,xn,y1,⋯,ym)=0(5)我们要求以上隐函数组能够确认一个隐向量函数
{
y
1
=
f
1
(
x
1
,
⋯
,
x
n
)
y
2
=
f
2
(
x
1
,
⋯
,
x
n
)
⋯
y
m
=
f
m
(
x
1
,
⋯
,
x
n
)
(6)
\tag{6} \begin{cases} y_1 = f_1(x_1,\cdots,x_n)\\ y_2 = f_2(x_1,\cdots,x_n)\\ \cdots\\ y_m = f_m(x_1,\cdots,x_n) \end{cases}
⎩⎪⎪⎪⎨⎪⎪⎪⎧y1=f1(x1,⋯,xn)y2=f2(x1,⋯,xn)⋯ym=fm(x1,⋯,xn)(6)满足什么条件的情况下,可以做到这一点呢?不妨假设
F
1
,
⋯
,
F
m
F_1,\cdots,F_m
F1,⋯,Fm在某个邻域上是连续可微的,可以确认一个隐函数组,并且每个分量函数都是连续可微的。那么求解隐函数组对各个分量的偏导数呢?以求解对
x
1
x_1
x1的偏导数为例:
将(6)代入(5),(5)的每个方程都对
x
1
x_1
x1求偏导,就有
{
∂
F
1
∂
y
1
∂
y
1
∂
x
1
+
∂
F
1
∂
y
2
∂
y
2
∂
x
1
+
⋯
+
∂
F
1
∂
y
m
∂
y
m
∂
x
1
=
−
∂
F
1
∂
x
1
∂
F
2
∂
y
1
∂
y
1
∂
x
1
+
∂
F
2
∂
y
2
∂
y
2
∂
x
1
+
⋯
+
∂
F
2
∂
y
m
∂
y
m
∂
x
1
=
−
∂
F
2
∂
x
1
⋯
∂
F
m
∂
y
1
∂
y
1
∂
x
1
+
∂
F
m
∂
y
2
∂
y
2
∂
x
1
+
⋯
+
∂
F
m
∂
y
m
∂
y
m
∂
x
1
=
−
∂
F
m
∂
x
1
(7)
\tag{7} \begin{cases} \frac{\partial F_1}{\partial y_1}\frac{\partial y_1}{\partial x_1} + \frac{\partial F_1}{\partial y_2}\frac{\partial y_2}{\partial x_1} + \cdots + \frac{\partial F_1}{\partial y_m}\frac{\partial y_m}{\partial x_1} =-\frac{\partial F_1}{\partial x_1}\\ \frac{\partial F_2}{\partial y_1}\frac{\partial y_1}{\partial x_1} + \frac{\partial F_2}{\partial y_2}\frac{\partial y_2}{\partial x_1} + \cdots + \frac{\partial F_2}{\partial y_m}\frac{\partial y_m}{\partial x_1} =-\frac{\partial F_2}{\partial x_1}\\ \cdots\\ \frac{\partial F_m}{\partial y_1}\frac{\partial y_1}{\partial x_1} + \frac{\partial F_m}{\partial y_2}\frac{\partial y_2}{\partial x_1} + \cdots + \frac{\partial F_m}{\partial y_m}\frac{\partial y_m}{\partial x_1} =-\frac{\partial F_m}{\partial x_1}\\ \end{cases}
⎩⎪⎪⎪⎪⎨⎪⎪⎪⎪⎧∂y1∂F1∂x1∂y1+∂y2∂F1∂x1∂y2+⋯+∂ym∂F1∂x1∂ym=−∂x1∂F1∂y1∂F2∂x1∂y1+∂y2∂F2∂x1∂y2+⋯+∂ym∂F2∂x1∂ym=−∂x1∂F2⋯∂y1∂Fm∂x1∂y1+∂y2∂Fm∂x1∂y2+⋯+∂ym∂Fm∂x1∂ym=−∂x1∂Fm(7)不难看出,(7)是关于
(
∂
y
1
∂
x
1
,
⋯
,
∂
y
m
∂
x
1
)
(\frac{\partial y_1}{\partial x_1},\cdots,\frac{\partial y_m}{\partial x_1})
(∂x1∂y1,⋯,∂x1∂ym)的齐次线性方程组,只要矩阵
J
=
[
∂
F
1
∂
y
1
∂
F
1
∂
y
2
⋯
∂
F
1
∂
y
m
∂
F
2
∂
y
1
∂
F
2
∂
y
2
⋯
∂
F
2
∂
y
m
⋯
⋯
⋯
⋯
∂
F
m
∂
y
1
∂
F
m
∂
y
2
⋯
∂
F
m
∂
y
m
]
J= \left[ \begin{matrix} \frac{\partial F_1}{\partial y_1}&\frac{\partial F_1}{\partial y_2}&\cdots&\frac{\partial F_1}{\partial y_m}\\ \frac{\partial F_2}{\partial y_1}&\frac{\partial F_2}{\partial y_2}&\cdots&\frac{\partial F_2}{\partial y_m}\\ \cdots&\cdots&\cdots&\cdots\\ \frac{\partial F_m}{\partial y_1}&\frac{\partial F_m}{\partial y_2}&\cdots&\frac{\partial F_m}{\partial y_m} \end{matrix} \right]
J=⎣⎢⎢⎢⎡∂y1∂F1∂y1∂F2⋯∂y1∂Fm∂y2∂F1∂y2∂F2⋯∂y2∂Fm⋯⋯⋯⋯∂ym∂F1∂ym∂F2⋯∂ym∂Fm⎦⎥⎥⎥⎤可逆,就可以反解出偏导数,可逆的充要条件是其行列式不为0,记
∂
(
F
1
,
⋯
,
F
m
)
∂
(
y
1
,
⋯
,
y
m
)
=
det
[
∂
F
1
∂
y
1
∂
F
1
∂
y
2
⋯
∂
F
1
∂
y
m
∂
F
2
∂
y
1
∂
F
2
∂
y
2
⋯
∂
F
2
∂
y
m
⋯
⋯
⋯
⋯
∂
F
m
∂
y
1
∂
F
m
∂
y
2
⋯
∂
F
m
∂
y
m
]
\frac{\partial(F_1,\cdots,F_m)}{\partial(y_1,\cdots,y_m)} =\det\left[ \begin{matrix} \frac{\partial F_1}{\partial y_1}&\frac{\partial F_1}{\partial y_2}&\cdots&\frac{\partial F_1}{\partial y_m}\\ \frac{\partial F_2}{\partial y_1}&\frac{\partial F_2}{\partial y_2}&\cdots&\frac{\partial F_2}{\partial y_m}\\ \cdots&\cdots&\cdots&\cdots\\ \frac{\partial F_m}{\partial y_1}&\frac{\partial F_m}{\partial y_2}&\cdots&\frac{\partial F_m}{\partial y_m} \end{matrix} \right]
∂(y1,⋯,ym)∂(F1,⋯,Fm)=det⎣⎢⎢⎢⎡∂y1∂F1∂y1∂F2⋯∂y1∂Fm∂y2∂F1∂y2∂F2⋯∂y2∂Fm⋯⋯⋯⋯∂ym∂F1∂ym∂F2⋯∂ym∂Fm⎦⎥⎥⎥⎤
定理14.8(隐函数存在定理2)
F
(
x
,
y
)
=
(
F
1
(
x
,
y
)
,
⋯
,
F
m
(
x
,
y
)
)
F(x,y)=(F_1(x,y),\cdots,F_m(x,y))
F(x,y)=(F1(x,y),⋯,Fm(x,y))是
R
n
+
m
R^{n+m}
Rn+m上的开集
G
G
G上的连续可微向量函数,其中
x
∈
R
n
,
y
∈
R
m
x\in R^n,y\in R^m
x∈Rn,y∈Rm,
(
x
0
,
y
0
)
∈
G
(x_0,y_0)\in G
(x0,y0)∈G,
x
0
=
(
x
1
0
,
⋯
,
x
n
0
)
,
y
0
=
(
y
1
0
,
⋯
,
y
m
0
)
x_0=(x_1^0,\cdots,x_n^0),y_0=(y_1^0,\cdots,y_m^0)
x0=(x10,⋯,xn0),y0=(y10,⋯,ym0),满足:
(1)
F
(
x
0
,
y
0
)
=
0
F(x_0,y_0)=0
F(x0,y0)=0
(2)
∂
(
F
1
,
⋯
,
F
m
)
∂
(
y
1
,
⋯
,
y
m
)
(
x
0
,
y
0
)
≠
0
\frac{\partial(F_1,\cdots,F_m)}{\partial(y_1,\cdots,y_m)}(x_0,y_0)\neq 0
∂(y1,⋯,ym)∂(F1,⋯,Fm)(x0,y0)=0
则存在
x
0
x_0
x0的邻域
B
1
B_1
B1及
y
0
y_0
y0的邻域
B
2
B_2
B2,
B
1
×
B
2
⊂
G
B_1\times B_2\subset G
B1×B2⊂G,存在惟一的
B
1
B_1
B1上的连续可微的
n
n
n元
m
m
m维向量函数
f
=
(
f
1
,
⋯
,
f
m
)
f=(f_1,\cdots,f_m)
f=(f1,⋯,fm),满足:
(1)
∀
x
∈
B
1
\forall x\in B_1
∀x∈B1,
(
x
,
f
(
x
)
)
∈
B
2
(x,f(x))\in B_2
(x,f(x))∈B2,
F
(
x
,
f
(
x
)
)
=
0
F(x,f(x))=0
F(x,f(x))=0
(2)
f
(
x
0
)
=
y
0
f(x_0)=y_0
f(x0)=y0
(3)
∀
x
∈
B
1
,
∂
(
F
1
,
⋯
,
F
m
)
∂
(
y
1
,
⋯
,
y
m
)
(
x
,
f
(
x
)
)
≠
0
\forall x\in B_1,\frac{\partial(F_1,\cdots,F_m)}{\partial(y_1,\cdots,y_m)}(x,f(x))\neq 0
∀x∈B1,∂(y1,⋯,ym)∂(F1,⋯,Fm)(x,f(x))=0,同时,令
D
(
x
,
y
)
=
[
∂
F
1
∂
y
1
(
x
,
y
)
∂
F
1
∂
y
2
(
x
,
y
)
⋯
∂
F
1
∂
y
m
(
x
,
y
)
∂
F
2
∂
y
1
(
x
,
y
)
∂
F
2
∂
y
2
(
x
,
y
)
⋯
∂
F
2
∂
y
m
(
x
,
y
)
⋯
⋯
⋯
∂
F
m
∂
y
1
(
x
,
y
)
∂
F
m
∂
y
2
(
x
,
y
)
⋯
∂
F
m
∂
y
m
(
x
,
y
)
]
D(x,y)=\left[\begin{matrix} \frac{\partial F_1}{\partial y_1}(x,y)&\frac{\partial F_1}{\partial y_2}(x,y)&\cdots&\frac{\partial F_1}{\partial y_m}(x,y)\\ \frac{\partial F_2}{\partial y_1}(x,y)&\frac{\partial F_2}{\partial y_2}(x,y)&\cdots&\frac{\partial F_2}{\partial y_m}(x,y)\\ \cdots&\cdots&&\cdots\\ \frac{\partial F_m}{\partial y_1}(x,y)&\frac{\partial F_m}{\partial y_2}(x,y)&\cdots&\frac{\partial F_m}{\partial y_m}(x,y) \end{matrix}\right]
D(x,y)=⎣⎢⎢⎢⎡∂y1∂F1(x,y)∂y1∂F2(x,y)⋯∂y1∂Fm(x,y)∂y2∂F1(x,y)∂y2∂F2(x,y)⋯∂y2∂Fm(x,y)⋯⋯⋯∂ym∂F1(x,y)∂ym∂F2(x,y)⋯∂ym∂Fm(x,y)⎦⎥⎥⎥⎤
A
(
x
,
y
)
=
[
∂
F
1
∂
x
1
(
x
,
y
)
∂
F
1
∂
x
2
(
x
,
y
)
⋯
∂
F
1
∂
x
n
(
x
,
y
)
∂
F
2
∂
x
1
(
x
,
y
)
∂
F
2
∂
x
2
(
x
,
y
)
⋯
∂
F
2
∂
x
n
(
x
,
y
)
⋯
⋯
⋯
∂
F
m
∂
x
1
(
x
,
y
)
∂
F
m
∂
x
2
(
x
,
y
)
⋯
∂
F
m
∂
x
n
(
x
,
y
)
]
A(x,y)=\left[\begin{matrix} \frac{\partial F_1}{\partial x_1}(x,y)&\frac{\partial F_1}{\partial x_2}(x,y)&\cdots&\frac{\partial F_1}{\partial x_n}(x,y)\\ \frac{\partial F_2}{\partial x_1}(x,y)&\frac{\partial F_2}{\partial x_2}(x,y)&\cdots&\frac{\partial F_2}{\partial x_n}(x,y)\\ \cdots&\cdots&&\cdots\\ \frac{\partial F_m}{\partial x_1}(x,y)&\frac{\partial F_m}{\partial x_2}(x,y)&\cdots&\frac{\partial F_m}{\partial x_n}(x,y) \end{matrix}\right]
A(x,y)=⎣⎢⎢⎡∂x1∂F1(x,y)∂x1∂F2(x,y)⋯∂x1∂Fm(x,y)∂x2∂F1(x,y)∂x2∂F2(x,y)⋯∂x2∂Fm(x,y)⋯⋯⋯∂xn∂F1(x,y)∂xn∂F2(x,y)⋯∂xn∂Fm(x,y)⎦⎥⎥⎤则
f
′
(
x
)
=
−
D
−
1
(
x
,
f
(
x
)
)
A
(
x
,
f
(
x
)
)
f^\prime(x)=-D^{-1}(x,f(x))A(x,f(x))
f′(x)=−D−1(x,f(x))A(x,f(x))
证:对隐函数组方程的个数进行归纳。在 m = 1 m=1 m=1时结论是成立的,假设对 k k k个方程确定的隐函数组,隐函数存在定理2也成立。
①对 k + 1 k+1 k+1个方程确定的隐函数组 F ( x , y ) = ( F 1 ( x , y ) , ⋯ , F k + 1 ( x , y ) ) = 0 F(x,y)=(F_1(x,y),\cdots,F_{k+1}(x,y))=0 F(x,y)=(F1(x,y),⋯,Fk+1(x,y))=0其中 x = ( x 1 , ⋯ , x n ) , y = ( y 1 , ⋯ , y k , y k + 1 ) x=(x_1,\cdots,x_n),y=(y_1,\cdots,y_k,y_{k+1}) x=(x1,⋯,xn),y=(y1,⋯,yk,yk+1), F ( x , y ) F(x,y) F(x,y)是 R n + k + 1 R^{n+k+1} Rn+k+1的开集 G G G上的连续可微映射,对 x 0 = ( x 1 0 , ⋯ , x n 0 ) , y 0 = ( y 1 0 , y 2 0 , ⋯ , y k + 1 0 ) x_0=(x_1^0,\cdots,x_n^0),y_0=(y_1^0,y_2^0,\cdots,y_{k+1}^0) x0=(x10,⋯,xn0),y0=(y10,y20,⋯,yk+10), ( x 0 , y 0 ) ∈ G (x_0,y_0)\in G (x0,y0)∈G,有
(1) F ( x 0 , y 0 ) = 0 F(x_0,y_0)=0 F(x0,y0)=0
(2) ∂ ( F 1 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ≠ 0 \frac{\partial(F_1,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{k+1})}(x_0,y_0)\neq0 ∂(y1,⋯,yk+1)∂(F1,⋯,Fk+1)(x0,y0)=0
存在 δ 1 > 0 \delta_1>0 δ1>0,当 ( x , y ) ∈ B ( ( x 0 , y 0 ) , δ 1 ) ⊂ G (x,y)\in B((x_0,y_0),\delta_1)\subset G (x,y)∈B((x0,y0),δ1)⊂G时,有 ∂ ( F 1 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y k + 1 ) ( x , y ) ≠ 0 \frac{\partial(F_1,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{k+1})}(x,y)\neq0 ∂(y1,⋯,yk+1)∂(F1,⋯,Fk+1)(x,y)=0,对行列式 det D ( x 0 , y 0 ) = ∂ ( F 1 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) \det D(x_0,y_0)=\frac{\partial(F_1,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{k+1})}(x_0,y_0) detD(x0,y0)=∂(y1,⋯,yk+1)∂(F1,⋯,Fk+1)(x0,y0)按第一行展开,就有 det D ( x 0 , y 0 ) = ∑ i = 1 k + 1 ( − 1 ) i + 1 ∂ F 1 ∂ x i ( x 0 , y 0 ) ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y i − 1 , y i + 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) \begin{aligned} &\det D(x_0,y_0)\\=&\sum_{i=1}^{k+1}(-1)^{i+1}\frac{\partial F_1}{\partial x_i}(x_0,y_0)\frac{\partial(F_{2},\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{i-1},y_{i+1},\cdots,y_{k+1})}(x_0,y_0) \end{aligned} =detD(x0,y0)i=1∑k+1(−1)i+1∂xi∂F1(x0,y0)∂(y1,⋯,yi−1,yi+1,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)则以上和式至少有一项非零,不失一般性,设 ∂ F 1 ∂ x 1 ( x 0 , y 0 ) ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ≠ 0 \frac{\partial F_1}{\partial x_1}(x_0,y_0)\frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x_0,y_0)\neq 0 ∂x1∂F1(x0,y0)∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)=0(否则调整 y 1 , ⋯ , y m y_1,\cdots,y_m y1,⋯,ym的顺序即可),则 ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ≠ 0 \frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x_0,y_0)\neq 0 ∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)=0,同时, F ( x 0 , y 0 ) = 0 F(x_0,y_0)=0 F(x0,y0)=0,由归纳假设:
令 y 0 ′ = ( y 2 0 , ⋯ , y k + 1 0 ) y_0^\prime=(y_2^0,\cdots,y_{k+1}^0) y0′=(y20,⋯,yk+10), x 0 ′ = ( x 1 0 , ⋯ , x n 0 , y 1 0 ) x_0^\prime=(x_1^0,\cdots,x_n^0,y_1^0) x0′=(x10,⋯,xn0,y10),存在 x 0 ′ x_0^\prime x0′的某个邻域 B 1 = B ( x 0 ′ , δ 2 ) B_1=B(x_0^\prime,\delta_2) B1=B(x0′,δ2)以及 y 0 ′ y_0^\prime y0′的某个邻域 B 2 = B ( y 0 ′ , δ 3 ) B_2=B(y_0^\prime,\delta_3) B2=B(y0′,δ3), B 1 × B 2 ⊂ G B_1\times B_2\subset G B1×B2⊂G,存在惟一的 B 1 B_1 B1上连续可微的 n + 1 n+1 n+1元 k k k维向量函数 φ ( x , y 1 ) = ( φ 2 ( x , y 1 ) , ⋯ , φ k + 1 ( x , y 1 ) ) \varphi(x,y_1)=(\varphi_2(x,y_1),\cdots,\varphi_{k+1}(x,y_1)) φ(x,y1)=(φ2(x,y1),⋯,φk+1(x,y1))。
(1) ∀ ( x , y 1 ) ∈ B 1 , F i ( x , y 1 , φ ( x , y 1 ) ) = 0 , i = 2. ⋯ , k + 1 , φ ( x , y 1 ) ∈ B 2 \forall (x,y_1)\in B_1,F_i(x,y_1,\varphi(x,y_1))=0,i=2.\cdots,k+1,\varphi(x,y_1)\in B_2 ∀(x,y1)∈B1,Fi(x,y1,φ(x,y1))=0,i=2.⋯,k+1,φ(x,y1)∈B2
(2) φ ( x 0 ′ ) = y 0 ′ \varphi(x_0^\prime)=y_0^\prime φ(x0′)=y0′
(3) ∀ ( x , y 1 ) ∈ B 1 \forall (x,y_1)\in B_1 ∀(x,y1)∈B1, ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x , y 1 , ϕ ( x , y 1 ) ) ≠ 0 \frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x,y_1,\phi(x,y_1))\neq0 ∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x,y1,ϕ(x,y1))=0,同时,令 D ′ ( x , y ) = [ ∂ F 2 ∂ y 2 ( x , y ) ⋯ ∂ F 2 ∂ y k + 1 ( x , y ) ⋯ ⋯ ∂ F k + 1 ∂ y 2 ( x , y ) ⋯ ∂ F k + 1 ∂ y k + 1 ( x , y ) ] D^\prime(x,y)=\left[\begin{matrix} \frac{\partial F_2}{\partial y_2}(x,y)&\cdots&\frac{\partial F_2}{\partial y_{k+1}}(x,y)\\ \cdots&&\cdots\\ \frac{\partial F_{k+1}}{\partial y_2}(x,y)&\cdots&\frac{\partial F_{k+1}}{\partial y_{k+1}}(x,y) \end{matrix}\right] D′(x,y)=⎣⎢⎡∂y2∂F2(x,y)⋯∂y2∂Fk+1(x,y)⋯⋯∂yk+1∂F2(x,y)⋯∂yk+1∂Fk+1(x,y)⎦⎥⎤ A ′ ( x , y ) = [ ∂ F 2 ∂ x 1 ( x , y ) ∂ F 2 ∂ x 2 ( x , y ) ⋯ ∂ F 2 ∂ x n ( x , y ) ∂ F 2 ∂ y 1 ( x , y ) ∂ F 3 ∂ x 1 ( x , y ) ∂ F 3 ∂ x 2 ( x , y ) ⋯ ∂ F 3 ∂ x n ( x , y ) ∂ F 3 ∂ y 1 ( x , y ) ⋯ ⋯ ⋯ ⋯ ∂ F k + 1 ∂ x 1 ( x , y ) ∂ F k + 1 ∂ x 2 ( x , y ) ⋯ ∂ F k + 1 ∂ x n ( x , y ) ∂ F k + 1 ∂ y 1 ( x , y ) ] A^\prime(x,y)=\left[\begin{matrix} \frac{\partial F_2}{\partial x_1}(x,y)&\frac{\partial F_2}{\partial x_2}(x,y)&\cdots&\frac{\partial F_2}{\partial x_n}(x,y)&\frac{\partial F_2}{\partial y_1}(x,y)\\ \frac{\partial F_3}{\partial x_1}(x,y)&\frac{\partial F_3}{\partial x_2}(x,y)&\cdots&\frac{\partial F_3}{\partial x_n}(x,y)&\frac{\partial F_3}{\partial y_1}(x,y)\\ \cdots&\cdots&&\cdots&\cdots\\ \frac{\partial F_{k+1}}{\partial x_1}(x,y)&\frac{\partial F_{k+1}}{\partial x_2}(x,y)&\cdots&\frac{\partial F_{k+1}}{\partial x_n}(x,y)&\frac{\partial F_{k+1}}{\partial y_1}(x,y) \end{matrix}\right] A′(x,y)=⎣⎢⎢⎢⎡∂x1∂F2(x,y)∂x1∂F3(x,y)⋯∂x1∂Fk+1(x,y)∂x2∂F2(x,y)∂x2∂F3(x,y)⋯∂x2∂Fk+1(x,y)⋯⋯⋯∂xn∂F2(x,y)∂xn∂F3(x,y)⋯∂xn∂Fk+1(x,y)∂y1∂F2(x,y)∂y1∂F3(x,y)⋯∂y1∂Fk+1(x,y)⎦⎥⎥⎥⎤就有 φ ′ = − D ′ ( x , y 1 , φ ( x , y 1 ) ) A ′ ( x , y 1 , φ ( x , y 1 ) ) \varphi^\prime=-D^\prime(x,y_1,\varphi(x,y_1))A^\prime(x,y_1,\varphi(x,y_1)) φ′=−D′(x,y1,φ(x,y1))A′(x,y1,φ(x,y1))这里规定 δ 2 < δ 1 , δ 3 < δ 1 \delta_2<\delta_1,\delta_3<\delta_1 δ2<δ1,δ3<δ1,这由连续性是可以取到的。
②我们要指出在 B 1 × B 2 B_1\times B_2 B1×B2上,方程 F ( x , y ) = 0 (I) \tag{I} F(x,y)=0 F(x,y)=0(I)和方程 { F 1 ( x , y ) = 0 y 2 = φ 2 ( x , y 1 ) ⋯ y k + 1 = φ k + 1 ( x , y 1 ) (II) \tag{II} \begin{cases} F_1(x,y)=0\\ y_2=\varphi_2(x,y_1)\\ \cdots\\ y_{k+1}=\varphi_{k+1}(x,y_1) \end{cases} ⎩⎪⎪⎪⎨⎪⎪⎪⎧F1(x,y)=0y2=φ2(x,y1)⋯yk+1=φk+1(x,y1)(II)是等价的。这是因为首先满足方程 ( I I ) (II) (II)自然就满足方程 ( I ) (I) (I),其次,如果满足 ( x , y ) (x,y) (x,y)方程 ( I ) (I) (I),并且 ( x , y 1 ) ∈ B 1 , ( y 2 , ⋯ , y k + 1 ) ∈ B 2 (x,y_1)\in B_1,(y_2,\cdots,y_{k+1})\in B_2 (x,y1)∈B1,(y2,⋯,yk+1)∈B2,则一定满足 F 1 ( x , y ) = 0 F_1(x,y)=0 F1(x,y)=0,其次还满足方程组 F i ( x , y ) = 0 , i = 2 , ⋯ , k + 1 F_i(x,y)=0,i=2,\cdots,k+1 Fi(x,y)=0,i=2,⋯,k+1。由于 ( x , y 1 ) ∈ B 1 , ( y 2 , ⋯ , y k + 1 ) ∈ B 2 (x,y_1)\in B_1,(y_2,\cdots,y_{k+1})\in B_2 (x,y1)∈B1,(y2,⋯,yk+1)∈B2,再由 φ \varphi φ的唯一性,就一定有 y i = φ i ( x , y 1 ) , i = 2 , ⋯ , k + 1 y_i=\varphi_i(x,y_1),i=2,\cdots,k+1 yi=φi(x,y1),i=2,⋯,k+1这就说明了 ( I ) (I) (I)和 ( I I ) (II) (II)是等价的。
③当 ( x , y 1 ) ∈ B 1 (x,y_1)\in B_1 (x,y1)∈B1时,令 G ( x , y 1 ) = F 1 ( x , y 1 , φ ( x , y 1 ) ) G(x,y_1)=F_1(x,y_1,\varphi(x,y_1)) G(x,y1)=F1(x,y1,φ(x,y1)),此时,由归纳假设, ( x , y 1 , φ ( x , y 1 ) ) ∈ B ( ( x 0 , y 0 ) , δ 1 ) (x,y_1,\varphi(x,y_1))\in B((x_0,y_0),\delta_1) (x,y1,φ(x,y1))∈B((x0,y0),δ1),则 ∂ g ∂ y 1 ( x 0 ′ ) = ∂ F 1 ∂ y 1 ( x 0 , y 0 ) + ∑ i = 2 k + 1 ∂ φ i ∂ y 1 ( x 0 ′ ) ∂ F 1 ∂ y i ( x 0 , y 0 ) \frac{\partial g}{\partial y_1}(x_0^\prime)=\frac{\partial F_1}{\partial y_1}(x_0,y_0)+\sum_{i=2}^{k+1}\frac{\partial \varphi_i}{\partial y_1}(x_0^\prime)\frac{\partial F_1}{\partial y_i}(x_0,y_0) ∂y1∂g(x0′)=∂y1∂F1(x0,y0)+i=2∑k+1∂y1∂φi(x0′)∂yi∂F1(x0,y0)同时 g ( x 0 ′ , φ ( x 0 ′ ) ) = F 1 ( x 0 , y 0 ) = 0 g(x_0^\prime,\varphi(x_0^\prime))=F_1(x_0,y_0)=0 g(x0′,φ(x0′))=F1(x0,y0)=0,现在我们来计算 ∂ g ∂ y 1 ( x 0 ′ ) \frac{\partial g}{\partial y_1}(x_0^\prime) ∂y1∂g(x0′)。由归纳假设, b = ( ∂ φ 2 ∂ y 1 ( x 0 ′ ) , ⋯ , ∂ φ k + 1 ∂ y 1 ( x 0 ′ ) ) T , a = ( ∂ F 2 ∂ y 1 ( x 0 , y 0 ) , ⋯ , ∂ F k + 1 ∂ y 1 ( x 0 , y 0 ) ) T b=(\frac{\partial \varphi_2}{\partial y_1}(x_0^\prime),\cdots,\frac{\partial \varphi_{k+1}}{\partial y_1}(x_0^\prime))^T,a=(\frac{\partial F_2}{\partial y_1}(x_0,y_0),\cdots,\frac{\partial F_{k+1}}{\partial y_1}(x_0,y_0))^T b=(∂y1∂φ2(x0′),⋯,∂y1∂φk+1(x0′))T,a=(∂y1∂F2(x0,y0),⋯,∂y1∂Fk+1(x0,y0))T,则有以下方程组成立 D ′ ( x 0 , y 0 ) b = − a D^\prime(x_0,y_0)b=-a D′(x0,y0)b=−a由克拉默法则,对 i = 2 , ⋯ , k + 1 i=2,\cdots,k+1 i=2,⋯,k+1,有 ∂ φ i ∂ y 1 ( x 0 ′ ) = ( − 1 ) i + 1 ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y i − 1 , y i + 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x 0 , y 0 ) \frac{\partial \varphi_i}{\partial y_1}(x_0^\prime)= \frac{ (-1)^{i+1}\frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{i-1},y_{i+1},\cdots,y_{k+1})}(x_0,y_0) }{ \frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x_0,y_0) } ∂y1∂φi(x0′)=∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)(−1)i+1∂(y1,⋯,yi−1,yi+1,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)则 ∂ g ∂ y 1 ( x 0 ′ ) = ∑ i = 1 k + 1 ( − 1 ) i + 1 ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y i − 1 , y i + 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x 0 , y 0 ) = ∂ ( F 1 , ⋯ , F k + 1 ) ∂ ( y 1 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ∂ ( F 2 , ⋯ , F k + 1 ) ∂ ( y 2 , ⋯ , y k + 1 ) ( x 0 , y 0 ) ≠ 0 \begin{aligned} &\frac{\partial g}{\partial y_1}(x_0^\prime)=\frac{ \sum_{i=1}^{k+1}(-1)^{i+1}\frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{i-1},y_{i+1},\cdots,y_{k+1})}(x_0,y_0) } { \frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x_0,y_0) }\\=&\frac{\frac{\partial(F_1,\cdots,F_{k+1})}{\partial(y_1,\cdots,y_{k+1})}(x_0,y_0)} {\frac{\partial(F_2,\cdots,F_{k+1})}{\partial(y_2,\cdots,y_{k+1})}(x_0,y_0)}\neq 0 \end{aligned} =∂y1∂g(x0′)=∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)∑i=1k+1(−1)i+1∂(y1,⋯,yi−1,yi+1,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)∂(y2,⋯,yk+1)∂(F2,⋯,Fk+1)(x0,y0)∂(y1,⋯,yk+1)∂(F1,⋯,Fk+1)(x0,y0)=0④由隐函数存在定理1,存在 x 0 x_0 x0的邻域 B 3 = B ( x 0 , δ 4 ) B_3=B(x_0,\delta_4) B3=B(x0,δ4)及 y 0 y_0 y0的邻域 B 4 = B ( y 0 , δ 5 ) B_4=B(y_0,\delta_5) B4=B(y0,δ5),存在唯一的定义在 B 3 B_3 B3上的 n n n元连续可微函数 y 1 = φ 1 ( x 1 , ⋯ , x n ) y_1=\varphi_1(x_1,\cdots,x_n) y1=φ1(x1,⋯,xn),满足:
(1) ∀ x ∈ B 3 , g ( x , φ 1 ( x ) ) = F 1 ( x , φ 1 ( x ) , φ ( x , φ 1 ( x ) ) ) = 0 , φ 1 ( x ) ∈ B 4 \forall x\in B_3,g(x,\varphi_1(x))=F_1(x,\varphi_1(x),\varphi(x,\varphi_1(x)))=0,\varphi_1(x)\in B_4 ∀x∈B3,g(x,φ1(x))=F1(x,φ1(x),φ(x,φ1(x)))=0,φ1(x)∈B4
(2) φ 1 ( x 0 ) = y 1 0 \varphi_1(x_0)=y_1^0 φ1(x0)=y10
(3) ∂ φ 1 ∂ x i ( x ) = − ∂ g ∂ x i ( x , φ 1 ( x ) ) ∂ g ∂ y 1 ( x , φ 1 ( x ) ) \frac{\partial\varphi_1}{\partial x_i}(x)=-\frac{\frac{\partial g}{\partial x_i}(x,\varphi_1(x))}{\frac{\partial g}{\partial y_1}(x,\varphi_1(x))} ∂xi∂φ1(x)=−∂y1∂g(x,φ1(x))∂xi∂g(x,φ1(x)), i = 1 , ⋯ , n , ∀ x ∈ B 3 i=1,\cdots,n,\forall x\in B_3 i=1,⋯,n,∀x∈B3
这里不妨设 δ 4 < δ 2 , δ 5 < min ( δ 2 , δ 3 ) \delta_4<\delta_2,\delta_5<\min(\delta_2,\delta_3) δ4<δ2,δ5<min(δ2,δ3)(由连续性是可以取到的)
⑤这里要指出的是,在 B 3 × B 4 ⊂ B 1 × B 2 B_3\times B_4 \subset B_1\times B_2 B3×B4⊂B1×B2上,方程组 ( I I ) (II) (II)和方程组 { y 1 = φ 1 ( x ) y 2 = φ 2 ( x , y 1 ) ⋯ y k + 1 = φ k + 1 ( x , y 1 ) (III) \tag{III} \begin{cases} y_1=\varphi_1(x)\\ y_2=\varphi_2(x,y_1)\\ \cdots\\ y_{k+1}=\varphi_{k+1}(x,y_1) \end{cases} ⎩⎪⎪⎪⎨⎪⎪⎪⎧y1=φ1(x)y2=φ2(x,y1)⋯yk+1=φk+1(x,y1)(III)是等价的,这说明在 B 3 × B 4 B_3\times B_4 B3×B4上, ( I I I ) (III) (III)和 ( I ) (I) (I)是等价的。
⑥对 x ∈ B 3 x\in B_3 x∈B3, ( x , φ 1 ( x ) ) ∈ B 1 (x,\varphi_1(x))\in B_1 (x,φ1(x))∈B1,令 { f 1 ( x ) = φ 1 ( x ) f 2 ( x ) = φ 2 ( x , φ 1 ( x ) ) ⋯ f k + 1 ( x ) = φ k + 1 ( x , φ 1 ( x ) ) \begin{cases} f_1(x)=\varphi_1(x)\\ f_2(x)=\varphi_2(x,\varphi_1(x))\\ \cdots\\ f_{k+1}(x)=\varphi_{k+1}(x,\varphi_1(x)) \end{cases} ⎩⎪⎪⎪⎨⎪⎪⎪⎧f1(x)=φ1(x)f2(x)=φ2(x,φ1(x))⋯fk+1(x)=φk+1(x,φ1(x)) f ( x ) = ( f 1 ( x ) , ⋯ , f k + 1 ( x ) ) f(x)=(f_1(x),\cdots,f_{k+1}(x)) f(x)=(f1(x),⋯,fk+1(x))在 B 3 B_3 B3上有定义且由构造 f ( x ) f(x) f(x)是连续可微的,同时 { f 1 ( x 0 ) = φ 1 ( x 0 ) = y 1 0 f 2 ( x 0 ) = φ 2 ( x 0 ′ ) = y 2 0 ⋯ f k + 1 ( x ) = φ k + 1 ( x 0 ′ ) = y k + 1 0 \begin{cases} f_1(x_0)=\varphi_1(x_0)=y_1^0\\ f_2(x_0)=\varphi_2(x_0^\prime)=y_2^0\\ \cdots\\ f_{k+1}(x)=\varphi_{k+1}(x_0^\prime)=y_{k+1}^0 \end{cases} ⎩⎪⎪⎪⎨⎪⎪⎪⎧f1(x0)=φ1(x0)=y10f2(x0)=φ2(x0′)=y20⋯fk+1(x)=φk+1(x0′)=yk+10并且对任意的 x ∈ B 3 x\in B_3 x∈B3,都有 { F 1 ( x , f ( x ) ) = g ( x , φ 1 ( x ) ) = 0 F 2 ( x , f ( x ) ) = F 2 ( x , φ 1 ( x ) , φ ( x , φ 1 ( x ) ) = 0 ⋯ F k + 1 ( x , f ( x ) ) = F k + 1 ( x , φ 1 ( x ) , φ ( x , φ 1 ( x ) ) = 0 \begin{cases} F_1(x,f(x))=g(x,\varphi_1(x))=0\\ F_2(x,f(x))=F_2(x,\varphi_1(x),\varphi(x,\varphi_1(x))=0\\ \cdots\\ F_{k+1}(x,f(x))=F_{k+1}(x,\varphi_1(x),\varphi(x,\varphi_1(x))=0 \end{cases} ⎩⎪⎪⎪⎨⎪⎪⎪⎧F1(x,f(x))=g(x,φ1(x))=0F2(x,f(x))=F2(x,φ1(x),φ(x,φ1(x))=0⋯Fk+1(x,f(x))=Fk+1(x,φ1(x),φ(x,φ1(x))=0此时 f ( x ) ∈ B 5 = B ( y 0 , min ( δ 3 , δ 4 ) ) < δ 1 f(x)\in B_5=B(y_0,\min(\delta_3,\delta_4))<\delta_1 f(x)∈B5=B(y0,min(δ3,δ4))<δ1, B 3 × B 5 ⊂ B 1 ⊂ G B_3\times B_5 \subset B_1 \subset G B3×B5⊂B1⊂G,偏导数只要左右两边求偏导,即可证得导数应当满足的方程组,并且由于 ( I ) (I) (I)和 ( I I I ) (III) (III)是等价的, f f f还是唯一的。由数学归纳法,定理14.8成立。
逆映射定理
为了阐述逆映射定理,首先要理解
微
分
同
胚
‾
\underline{微分同胚}
微分同胚和
局
部
微
分
同
胚
‾
\underline{局部微分同胚}
局部微分同胚的概念,首先要理解何谓同胚。
X
⊂
R
n
,
Y
⊂
R
m
X\subset R^n,Y\subset R^m
X⊂Rn,Y⊂Rm,
ϕ
:
X
→
Y
\phi:X\to Y
ϕ:X→Y是
n
n
n元
m
m
m维向量函数
(1)如果
∀
x
1
,
x
2
∈
X
\forall x_1,x_2\in X
∀x1,x2∈X,
ϕ
(
x
1
)
≠
ϕ
(
x
2
)
\phi(x_1)\neq \phi(x_2)
ϕ(x1)=ϕ(x2),则称
ϕ
\phi
ϕ是单射
(2)如果
ϕ
(
X
)
=
Y
\phi(X)=Y
ϕ(X)=Y,则称
ϕ
\phi
ϕ为满射
如果
ϕ
\phi
ϕ既是单射,又是满射,那么,存在逆映射
φ
:
Y
→
X
\varphi:Y\to X
φ:Y→X,该逆映射同时满足:
(1)
ϕ
(
φ
(
y
)
)
=
y
∀
y
∈
Y
\phi(\varphi(y))=y\quad\forall y\in Y
ϕ(φ(y))=y∀y∈Y
(2)
φ
(
ϕ
(
x
)
)
=
x
∀
x
∈
X
\varphi(\phi(x))=x\quad\forall x\in X
φ(ϕ(x))=x∀x∈X
如果
ϕ
\phi
ϕ是连续的,
φ
\varphi
φ也是连续的,则称
ϕ
\phi
ϕ是
X
X
X到
Y
Y
Y上的同胚映射,如果
ϕ
\phi
ϕ是连续可微的,
φ
\varphi
φ也是连续可微的,则称
ϕ
\phi
ϕ是
X
X
X到
Y
Y
Y的微分同胚映射。那么何谓局部微分同胚呢?
Ω
⊂
R
n
\Omega\subset R^n
Ω⊂Rn是开集,
Y
⊂
R
m
Y\subset R^m
Y⊂Rm,
ϕ
:
Ω
→
Y
\phi:\Omega\to Y
ϕ:Ω→Y,对
x
0
∈
Ω
x_0\in \Omega
x0∈Ω,如果存在包含
x
0
x_0
x0的一个开集
G
1
⊂
Ω
G_1\subset \Omega
G1⊂Ω,以及包含
y
0
y_0
y0的一个开集
G
2
⊂
Y
G_2\subset Y
G2⊂Y,使得
ϕ
\phi
ϕ是
G
1
G_1
G1到
G
2
G_2
G2的微分同胚映射,则称
ϕ
\phi
ϕ在
x
0
x_0
x0处局部微分同胚。于是,我们可以给出逆映射定理。
定理14.9 ϕ : Ω → R n \phi:\Omega\to R^n ϕ:Ω→Rn是开集 Ω ⊂ R n \Omega\subset R^n Ω⊂Rn上的连续可微映射, x 0 ∈ Ω x_0\in \Omega x0∈Ω,如果 det D ϕ ( x 0 ) ≠ 0 \det D\phi(x_0)\neq 0 detDϕ(x0)=0,则 ϕ \phi ϕ在 x 0 x_0 x0处局部微分同胚
我们先来整理定理的结论:所谓
ϕ
\phi
ϕ在
x
0
x_0
x0处微分同胚,有两层含义:
(1)
ϕ
\phi
ϕ在包含
x
0
x_0
x0的某个开集
G
1
(
G
1
⊂
Ω
)
G_1(G_1\subset \Omega)
G1(G1⊂Ω)上存在逆映射
(2)
ϕ
(
G
1
)
\phi(G_1)
ϕ(G1)也是开集,并且
ϕ
,
ϕ
−
1
\phi,\phi^{-1}
ϕ,ϕ−1都是连续可微的
因此,我们只需要证明这两点即可,而证明第二点用到的就是隐函数存在定理。
证:
令 F i ( x , y ) = y i − φ i ( x ) ( i = 1 , ⋯ , n ) F_i(x,y)=y_i-\varphi_i(x)(i=1,\cdots,n) Fi(x,y)=yi−φi(x)(i=1,⋯,n),令 F ( x , y ) = ( F 1 ( x , y ) , ⋯ , F n ( x , y ) ) F(x,y)=(F_1(x,y),\cdots,F_n(x,y)) F(x,y)=(F1(x,y),⋯,Fn(x,y)), y 0 = φ ( x 0 ) y_0=\varphi(x_0) y0=φ(x0),有 F ( x 0 , y 0 ) = 0 F(x_0,y_0)=0 F(x0,y0)=0,同时 ∂ ( F 1 , ⋯ , F n ) ∂ ( x 1 , ⋯ , x n ) ( x 0 , y 0 ) = det D φ ( x 0 ) ≠ 0 \frac{\partial(F_1,\cdots,F_n)}{\partial(x_1,\cdots,x_n)}(x_0,y_0)=\det D \varphi(x_0)\neq 0 ∂(x1,⋯,xn)∂(F1,⋯,Fn)(x0,y0)=detDφ(x0)=0由定理14.8,存在 y 0 y_0 y0的邻域 B 1 = B ( y 0 , δ 1 ) B_1=B(y_0,\delta_1) B1=B(y0,δ1)及 x 0 x_0 x0的邻域 B 2 = B ( x 0 , δ 2 ) ⊂ G B_2=B(x_0,\delta_2)\subset G B2=B(x0,δ2)⊂G,存在唯一的定义在 B 1 B_1 B1的连续可微的 n n n元 n n n维向量函数 x = ϕ ( y ) x=\phi(y) x=ϕ(y),满足:
(1) ∀ y ∈ B 1 F ( ϕ ( y ) , y ) = 0 , ϕ ( y ) ∈ B 2 \forall y \in B_1\quad F(\phi(y),y)=0,\phi(y)\in B_2 ∀y∈B1F(ϕ(y),y)=0,ϕ(y)∈B2
(2) ϕ ( y 0 ) = x 0 \phi(y_0)=x_0 ϕ(y0)=x0
(3) D ϕ ( y ) = D − 1 φ ( ϕ ( y ) ) ∀ y ∈ B 1 D\phi (y)=D^{-1}\varphi(\phi(y))\quad \forall y\in B_1 Dϕ(y)=D−1φ(ϕ(y))∀y∈B1
令 S = { x ∈ B 2 : φ ( x ) ∈ B 1 } S=\{x\in B_2:\varphi(x)\in B_1\} S={x∈B2:φ(x)∈B1},下面证明 S S S是一个开集,对任意的 x ∈ S x\in S x∈S, B 2 B_2 B2是开集,存在 δ 3 > 0 \delta_3>0 δ3>0,有 B ( x , δ 3 ) ⊂ B 2 B(x,\delta_3)\subset B_2 B(x,δ3)⊂B2 φ ( x ) ∈ B 1 \varphi(x)\in B_1 φ(x)∈B1,由 B 1 B_1 B1是开集,存在 δ 4 > 0 \delta_4>0 δ4>0,有 B ( φ ( x ) , δ 4 ) ⊂ B 1 B(\varphi(x),\delta_4)\subset B_1 B(φ(x),δ4)⊂B1再由 φ \varphi φ的连续性,存在 δ 5 < δ 3 \delta_5<\delta_3 δ5<δ3,满足:当 x ′ ∈ B ( x , δ 5 ) x^\prime\in B(x,\delta_5) x′∈B(x,δ5)时,都有 φ ( x ′ ) ∈ B ( φ ( x ) , δ 4 ) \varphi(x^\prime) \in B(\varphi(x),\delta_4) φ(x′)∈B(φ(x),δ4)因此 B ( x , δ 5 ) ⊂ S B(x,\delta_5) \subset S B(x,δ5)⊂S从而 S S S是开集,同时 φ ( S ) = B 1 \varphi(S)=B_1 φ(S)=B1, φ \varphi φ是 S S S到 B 1 B_1 B1的连续可微映射,由 ϕ \phi ϕ的唯一性, φ \varphi φ是单射,同时, ϕ \phi ϕ是 φ \varphi φ的逆映射,由隐函数存在定理,两者均连续可微,因此, φ \varphi φ在 x 0 x_0 x0处局部微分同胚。
定理14.10 φ : R n → R n \varphi:R^n\to R^n φ:Rn→Rn是开集 G G G上的连续可微映射,并且 det D φ ( x ) ≠ 0 ∀ x ∈ G \det D\varphi(x)\neq 0\quad \forall x\in G detDφ(x)=0∀x∈G,则 φ \varphi φ将 G G G的任意开子集映成开集
证:
对任意的开集 Ω ⊂ G \Omega\subset G Ω⊂G,对任意的 x ∈ Ω x\in \Omega x∈Ω,存在包含 x x x的开集 Ω x ⊂ Ω \Omega_x\subset \Omega Ωx⊂Ω,使得 φ \varphi φ是 Ω x \Omega_x Ωx到 φ ( Ω x ) \varphi(\Omega_x) φ(Ωx)的微分同胚映射, φ ( Ω x ) \varphi(\Omega_x) φ(Ωx)是开集。由于 Ω = ⋃ x ∈ Ω Ω x \Omega=\bigcup_{x\in \Omega}\Omega_x Ω=x∈Ω⋃Ωx因此 φ ( Ω ) = ⋃ x ∈ Ω φ ( Ω x ) \varphi(\Omega)=\bigcup_{x\in \Omega}\varphi(\Omega_x) φ(Ω)=x∈Ω⋃φ(Ωx)由于开集的任意并是开集,因此, φ ( Ω ) \varphi(\Omega) φ(Ω)是开集。
虽然向量函数的Jocobi矩阵起到了导数的作用,在一维情形下,如果连续可微,且导数在开区间上不为0,就可以推出函数是严格单调的,从而存在逆映射,但在高维空间上并非如此。
例14.2 雅克比行列式为0推不出单射 φ ( x , y ) : { u = x 2 − y 2 v = 2 x y \varphi(x,y):\begin{cases} u=x^2-y^2\\ v=2xy \end{cases} φ(x,y):{u=x2−y2v=2xy Ω = R 2 / { ( 0 , 0 ) } \Omega=R^2/\{(0,0)\} Ω=R2/{(0,0)},则 φ \varphi φ在 Ω \Omega Ω上的雅克比行列式恒不为0 det D φ ( x , y ) = 4 ( x 2 + y 2 ) > 0 ∀ ( x , y ) ≠ ( 0 , 0 ) \det D\varphi(x,y)=4(x^2+y^2)>0\quad\forall (x,y)\neq (0,0) detDφ(x,y)=4(x2+y2)>0∀(x,y)=(0,0)但 φ ( x , y ) = φ ( − x , − y ) \varphi(x,y)=\varphi(-x,-y) φ(x,y)=φ(−x,−y),从而 φ ( x , y ) \varphi(x,y) φ(x,y)在 Ω \Omega Ω上不是单射。