next up previous
Next: 極大和極小 Up: 多變數函數的微分學 Previous: Tangent Spaces and Differentials

Higher Order Derivatives

If $ \displaystyle\frac{\partial f}{\partial x_{j} } $ is difinied, it is also a real valued function of n variables. The partial derivatives of $ \displaystyle\frac{\partial f}{\partial x_{1} } ,\ldots , \frac{\partial f}{\partial x_{n}} $ are called second order partial derivatives (二階偏導函數) of f. It is customary to write $ \displaystyle\frac{\partial^{2} f}{\partial x_{j}\partial x_{k}}$ or $ \displaystyle f_{x_{k} x_{j} } $ for $ \displaystyle\frac{\partial}{\partial x_{j} } \left( \frac{\partial f}{\partial x_{k} } \right) $. Derivatives of orders higher than 2 are definied similarly. Thus,

\begin{displaymath}\frac{\partial^{m} f}{\partial x_{k_{1}} \partial x_{k_{2}} \...
...1true cm{\hfill} f_{x_{k_{m} }x_{k_{m-1} }\cdots x_{k_{1} } }
\end{displaymath}

both denote the m th order partial derivative $ \displaystyle\frac{\partial}{\partial x_{k_{1} } } \cdot \frac{\partial}{\partial x_{k_{2} } } \cdots \left( \frac{\partial f}{\partial x_{k_{m} } }\right) $.

我們舉兩個簡單的例子如下:

例 9   令 $ f(x,y)=-\cos(x^{2},y) $. 則

\begin{displaymath}f_{x} =2xy\sin(x^{2}y), \hbox to 1.5truecm{\hfill} f_{y}=sin(x^{2}y).
\end{displaymath}

於是

\begin{eqnarray*}f_{xx} & = & 2y\sin{(x^{2} y)}+4x^{2} y^{2}\cos{(x^{2}y)}, \\
...
...+2x^{3} y\cos{(x^{2}y)}, \\
f_{yy} & = & x^{4}\cos{(x^{2}y)}.
\end{eqnarray*}


例 10   令 f(x, y, z)=x3y2z. 則 fxy=fyx=6xyz, 且 fxyz, fxzy, fyzx, fyxz, fzxy, fzyx 均等於 6x2y.

在這兩個例子裡都有 fxy=fyx, 而且關於不同變數的混合偏微分均似與變數的次序無關. 這項觀察大體不錯, 因為我們有下定理:


\begin{theorem}Let $ f(x,y) $\space be a function defined in a domain $ D $\spac...
...in{displaymath}
f_{xy}(a,b)=f_{yx}(a,b).
\end{displaymath}\space
\end{theorem}

證明. 選擇 (x,y) 使 $x\neq a$, $y\neq b$. 令

Q=f(x,y)-f(a,y)-f(x,b)+f(a,b)

先暫時固定 x, 定義 g(y)=f(x,y)-f(a,y). 則由均值定理得

\begin{displaymath}Q=g(y)-g(b)=g'(\eta)(y-b)=[f_{y}(x,\eta )-f_{y }(a,\eta )](y-b)
\end{displaymath}

式中 $ \eta $ 之值在 yb 之間. 現在把 x 當成變數, 再用一次均值定理, 乃得

\begin{displaymath}Q=f_{yx} (\xi ,\eta )(x-a)(y-b)
\end{displaymath}

式中 $\xi$ 之值在 xa 之間.

仿此先固定 y, 定義 h(x)=f(x, y)-f(x, b). 引用均值定理兩次乃得

\begin{displaymath}Q=f_{xy}(\xi ', \eta ')(x-a)(y-b),
\end{displaymath}

式中 $ \xi'$$\eta'$ 之值分別在 x, ay, b 之間. 比較兩個 Q 式, 遂有 $ f_{yx} (\xi,\eta )=f_{xy}(\xi ', \eta '). $ $ (x,y) \rightarrow (a,b) $, 由 fxyfyx 的連續性知

\begin{displaymath}f_{yx} (a,b)=\lim_{(x,y) \rightarrow (a,b)} f(\xi,\eta)=\lim_{(x,y) \rightarrow (a, b)} f_{xy} f(\xi' , \eta ')=f_{xy}(a,b).
\end{displaymath}

明所欲證.

從以上的討論我們大體上可以相信 fyx=fxy 一般都成立了. 但事實上也有二者不相等的例子:

例 11   令

\begin{displaymath}f(x,y)=\left\{ \begin{array}{ll}
\displaystyle\frac{xy(x^{2}-...
...{\smallskip }
0 & \mbox{if } (x, y)=(0, 0)
\end{array}\right.
\end{displaymath}

易證 f 在整個平面上連續. 若 $ (x, y)\neq(0, 0) $

\begin{displaymath}f_{x}(x, y)=\frac{4x^{2}y^{3}+x^{4}y-y^{5}}{(x^{2}+y^{2})^{2}},\hbox to 1truecm{\hfill} f_{x}(0, 0)=0
\end{displaymath}

從而知 fx 在整個平面上連續. 因 f(y, x)=-f(x, y), 故有

\begin{displaymath}f_{y}(x, y)=-f_{y}(y, x)=-\frac{4x^{2}y^{3}+x^{4}y-x^{5}}{(x^{2}+y^{2})^{2}}, \hbox to 1truecm{\hfill} f_{y}(0, 0)=0
\end{displaymath}

fx 也在整個平面上連續. 由以上的結果很容易算出

\begin{displaymath}f_{xy}(0, 0)=-1,\quad f_{yx}(0, 0)=1 . \end{displaymath}

$\quad$
1.
Show that the minimal distance from a point $ y\in \mathbb{R} ^n $to the hyperplane through a point p with an unit normal vector u is given by $ \vert (y-p)\cdot u \vert $.

2.
A curve which satisfies an equation of the form

Ax2+Bxy+Cy2+Dx+Ey+F=0.

is called a conic. Assume that this conic passes through the point (a,b). Show that the tangent line to this conic at (a,b) is given by

\begin{displaymath}Aax+B\frac {bx+ay}{2}+Cby+D\frac {x+a}{2}+E\frac {y+b}{2}+F=0.
\end{displaymath}

3.
Let $ u\in \mathbb{R} ^n, \prod=\{v\in \mathbb{R} ^n:u\cdot v=0\}$. If $ w\cdot v=0 $ for all $ v\in \prod $, show that $w=\lambda u$ for some $\lambda \in \mathbb{R} $.

4.
Show that every tangent plane to cone

z2=x2+y2.

intersects the cone in a straight line.

5.
試計算下列三函數的各二階偏導函數:(a) $\displaystyle\sin(x-y)+\sin(x+y)$,     (b) $\displaystyle\arcsin \frac {x}{\sqrt {x^2+y^2}}$,     (c) $\displaystyle\cosh\ln(xy)$.

6.
Let u=x2siny+y2sinx. Find $ \displaystyle\frac {\partial ^6u}{\partial x^3\partial y^3}$.

7.
$ (x, y)\neq(0, 0) $. 令 $ u=ln \sqrt{x^2+y^2} $. 試證

\begin{displaymath}\frac {\partial ^2u}{\partial x^2}+\frac {\partial ^2u}{\partial y^2}=0.
\end{displaymath}

8.
$ (x,y,z)\neq (0,0,0) $. 令 $ u=\frac {1}{\sqrt{x^2+y^2+z^2} } $. 試證

\begin{displaymath}\frac {\partial ^2u}{\partial x^2}+\frac {\partial ^2u}{\partial y^2}+\frac {\partial ^2u}{\partial z^2}=0.
\end{displaymath}

9.
定義函數 $ u(x,t)=e^{-an^2t}\sin \,nx $, 式中 an 為常數. 試證明 u 滿足微分方程式 ut=auxx.

10.
(a) Let F(t)=f(x,y) where $ f(x,y)=e^x\cos y$, x=t2, y=1-t3. Compute F'(t) and F''(t) first by substitution and then by the chain rule.
(b) Do the same thing with f(x,y,z)=ex2+y2+z2, $x=\cos t$, $y=\sin t$, z=t.

11.
$ u=u(x,y), x=r\cos \theta, y=r\sin \theta $, 試證明

\begin{displaymath}u_{xx}+u_{yy}=u_{rr}+\frac {1}{r}u_r+\frac {1}{r^2}u_{\theta \theta}.
\end{displaymath}

12.
u=u(x, y, z), $x=\rho \cos\theta \cos\varphi$, $y=\rho \sin\theta \cos\varphi$, $z=\rho \sin\varphi$, 試將 uxx+ uyy+uzz 用關於 $\rho$, $\theta$$\varphi$ 的偏導函數表出.

next up previous
Next: 極大和極小 Up: 多變數函數的微分學 Previous: Tangent Spaces and Differentials

1999-06-28