8.2.Jacobi矩阵与全微分
2.1.向量集映射与Jacobi矩阵
考虑向量集函数\(\mathbf{z}=f(\mathbf{v}),\mathbf{v}=(x,y)^{\mathbf{T}}\)在\((x_0,y_0)\)处微分:
\[
f(\mathbf v+\mathbf h)-f(\mathbf v)=J\cdot\mathbf h+o(||\mathbf h||),\mathbf h \to \mathbf 0
\]
其中\(o(||\mathbf h||)\)为\(||\mathbf h||\)的高阶无穷小,因此\(f(\mathbf v)\)在\((x_0,y_0)\)处任意方向\(\mathbf h\)的增量可近似于增量\(\mathbf h\)的线性函数:
\[
\text{d}f(\mathbf v)=J\cdot\mathbf h
\]
对于更为具体的向量集函数\(F(x,y)=(P(x,y),Q(x,y))^{\mathbf T}\),其微分可表示为:
\[
\text{d}F(x,y)=
\begin{pmatrix}
\text{d}P\\
\text{d}Q
\end{pmatrix}
=\begin{pmatrix}
\nabla P\\
\nabla Q
\end{pmatrix}\cdot\mathbf h=
\begin{pmatrix}
P_x & P_y\\
Q_x & Q_y
\end{pmatrix}\cdot\mathbf{h}
\]
因此我们确定二维向量集函数\(\mathbf{z}=f(\mathbf{v})\)的Jacobi矩阵:
\[
J=\begin{pmatrix}
P_x & P_y\\
Q_x & Q_y
\end{pmatrix}
\]
对于二维值函数\(z=f(x,y)\),有\(\text{d}z=\nabla f\cdot\mathbf{h}\),\(\nabla f\)为值函数\(z=f(x,y)\)的Jacobi矩阵,被视为一个二维向量向量到值的映射。
2.2.链式法则的Jacobi矩阵形式
对于二维的复合值函数\(z=F(u(x,y),v(x,y))\),其微分为:
\[
\text{d}z=\nabla F\cdot
\begin{pmatrix}
\text{d}u \\
\text{d}v
\end{pmatrix}=
\nabla F \cdot
\begin{pmatrix}
\nabla u \\
\nabla v
\end{pmatrix}\cdot
\begin{pmatrix}
\text{d}x \\
\text{d}y
\end{pmatrix}=
\nabla F\cdot
\begin{pmatrix}
u_x & u_y\\
v_x & v_y
\end{pmatrix}\cdot
\begin{pmatrix}
\text{d}x \\
\text{d}y
\end{pmatrix}
\]
对于更为普遍的复合函数,链式法则展示了一种\(n\)维向量通过对应Jacobi矩阵的一种链式映射关系。
如上述的复合函数\(z=F(\lambda(x,y),\mu(x,y))\),可视为下述的链式映射:
\[
\begin{pmatrix}
\text{d}x\\
\text{d}y
\end{pmatrix}
\stackrel{J_{(u,v)}}{\longrightarrow}
\begin{pmatrix}
\text{d}u\\
\text{d}v
\end{pmatrix}
\stackrel{J_F}{\longrightarrow}
\text{d}z
\]
2.3.隐函数与隐映射
对于隐函数\(f(x_1,x_2,x_3)=0\),可以确定唯一的映射:\(x_i=x_i(x_j,x_k)\)。对函数\(f\)微分:
\[
\text{d}f=\nabla f\cdot
\begin{pmatrix}
\nabla x_1 \\
\nabla x_2 \\
\nabla x_3
\end{pmatrix}\cdot
\begin{pmatrix}
\text{d}x_1 \\
\text{d}x_2
\end{pmatrix}=
\nabla f\cdot
\begin{pmatrix}
I\\
\nabla x_3
\end{pmatrix}\cdot
\begin{pmatrix}
\text{d}x_1 \\
\text{d}x_2
\end{pmatrix}
\]
解出\(\nabla x_3\):
\[
\nabla f_{(x_1,x_2)}+\frac{\partial f}{\partial x_3}\nabla x_3=0
\]
即:\(\nabla x_3=-{(\frac{\partial f}{\partial x_3})}^{-1}\nabla f_{(x_1,x_2)}\)。
对于隐函数方程组:
\[
\left\{
\begin{matrix}
F_1(x,y,u,v)=0 \\
F_2(x,y,u,v)=0
\end{matrix}
\right.
\]
可以确定唯一隐映射:
\[
\left\{
\begin{matrix}
u = u(x,y) \\
v = v(x,y)
\end{matrix}
\right.
\]
对\(F_1,F_2\)微分:
\[
\left\{
\begin{matrix}
\text{d}F_1=\nabla F_1\cdot J_{(u,v)}\cdot\mathbf{h} \\
\text{d}F_2=\nabla F_2\cdot J_{(u,v)}\cdot\mathbf{h}
\end{matrix}
\right.
\]
不妨写作以下形式:
\[
\text{d}\mathbf{F}=
\begin{pmatrix}
\text{d}F_1 \\
\text{d}F_2
\end{pmatrix}=
\begin{pmatrix}
\nabla F_1 \\
\nabla F_2
\end{pmatrix}\cdot
J_{(u,v)}\cdot\mathbf{h}=
J_F\cdot J_{(u,v)} \cdot \mathbf{h}
\]
由于微分的形式不变性,对\(\mathbf F\)的微分依然可以视作对应Jacobi矩阵的链式映射法则:
\[
\begin{pmatrix}
\text{d}x\\
\text{d}y
\end{pmatrix}
\stackrel{J_{(u,v)}}{\longrightarrow}
\begin{pmatrix}
\text{d}\lambda\\
\text{d}\mu
\end{pmatrix}
\stackrel{J_F}{\longrightarrow}
\text{d}\mathbf F
\]
不妨将\(J_F\)分块为:\((J_{F_{(x,y)}}\ \ \ J_{F_{(u,v)}}),|J_F|\ne 0\),有:
\[
\text{d}\mathbf{F}=
\begin{pmatrix}
J_{F_{(x,y)}} & J_{F_{(u,v)}}
\end{pmatrix}\cdot
\begin{pmatrix}
I \\
J_{(u,v)}
\end{pmatrix}\cdot
\mathbf{h}=
(J_{F_{(x,y)}}+J_{F_{(u,v)}}J_{(u,v)})\cdot\mathbf{h}
\]
解出\(J_{(\lambda,\mu)}\):
\[
J_{(u,v)}=-{(J_{F_{(u,v)}})}^{-1}J_{F_{(x,y)}}
\]
2.4.曲面参数化与Jaccobi行列式
对于光滑且连续的曲面\(\Omega\),其自变量\(x,y\)可作参数化表示:\(x=P(u,v),y=Q(u,v)\),可视作\(x,y\)关于\(u,v\)的向量场\(\mathbf{z}=\mathbf{F}(u,v)\)。对于\(\text{x}o\text{y}\)平面上任意一点,\(J\)可视作对该点处邻域的线性变换。即:\(\text{d}\mathbf{z}=J\text{d}\mathbf{h}\),此处面积微元:
\[
\text{d}\sigma=\text{d}x\text{d}y=\text{det}J\cdot\text{d}u\text{d}v
\]
- 极坐标
\[
\left\{
\begin{matrix}
x = r\cos\theta \\
y = r\sin\theta
\end{matrix}
\right.
\]
可表示为:\(\text{d}\mathbf{z}=J\text{d}\mathbf{h}(r,\theta)\)。其中:
\[
J=\begin{pmatrix}
\cos\theta & -r\sin\theta\\
\sin\theta & r\cos\theta
\end{pmatrix}, \text{det}J=r
\]
故有:
\[
\text{d}\sigma=\text{d}x\text{d}y=r\text{d}r\text{d}\theta
\]
- 球坐标
\[
\left\{
\begin{matrix}
x = r\sin\theta\cos\phi \\
y = r\sin\theta\sin\phi \\
z = r\cos\theta
\end{matrix}
\right.
\]
可表示为:\(\text{d}\mathbf{\omega}=J\text{d}\mathbf{h}(r,\theta,\phi)\)。其中:
\[
J=\begin{pmatrix}
\sin\theta\cos\phi & r\cos\theta\cos\phi & -r\sin\theta\sin\phi\\
\sin\theta\sin\phi & r\cos\theta\sin\phi & r\sin\theta\cos\phi\\
\cos\theta & -r\sin\theta & 0
\end{pmatrix}, \text{det}J=r^2\sin\theta
\]
故有:
\[
\text{d}V=\text{d}x\text{d}y\text{d}z=r^2\sin\theta\text{d}r\text{d}\theta\text{d}\phi
\]
- 柱坐标
\[
\left\{
\begin{matrix}
x = r\cos\theta \\
y = r\sin\theta \\
z = z
\end{matrix}
\right.
\]
可表示为:\(\text{d}\mathbf{\omega}=J\text{d}\mathbf{h}(r,\theta, z)\)。其中:
\[
J=\begin{pmatrix}
\cos\theta & -r\sin\theta & 0\\
\sin\theta & r\cos\theta & 0\\
0 & 0 & 1
\end{pmatrix}, \text{det}J=r
\]
故有:
\[
\text{d}V=\text{d}x\text{d}y\text{d}z=r\text{d}r\text{d}\theta\text{d}z
\]
- 锥面坐标
\[
\left\{
\begin{matrix}
x = r\cos\theta \\
y = r\sin\theta \\
z = kr
\end{matrix}
\right.
\]
可表示为:\(\text{d}\mathbf{\omega}=J\text{d}\mathbf{h}(r,\theta, k)\)。其中:
\[
J=\begin{pmatrix}
\cos\theta & -r\sin\theta & 0\\
\sin\theta & r\cos\theta & 0\\
k & 0 & r
\end{pmatrix}, \text{det}J=r^2
\]
故有:
\[
\text{d}V=\text{d}x\text{d}y\text{d}z=r^2\text{d}r\text{d}\theta\text{d}k
\]