475 lines
23 KiB
TeX
475 lines
23 KiB
TeX
\chapter{多重积分}
|
||
\section{矩形区域上的积分}
|
||
\begin{definition}[二重积分]
|
||
设$I$是$\realnum^2$中的闭矩形，$I = [a, b] \times [c, d]$。$f: I \to \realnum$。作$[a, b]$的分割
|
||
\[\pi_x: a = x_0 < x_1 < \dots < x_n = b\]
|
||
又作$[c, d]$的分割
|
||
\[\pi_y: c = y_0 < y_1 < \dots < y_m = d\]
|
||
两组平行线把$I$分割成$k = n \times m$个子矩形
|
||
\[[x_{i - 1}, x_i] \times [y_{j - 1}, y_j], i = 1, 2, \dots, n, j = 1, 2, \dots, m\]
|
||
这$k$个子矩形的全体组成$I$的一个分割$\pi = \pi_x \times \pi_y$，用一定的次序重排这$k$个子矩形，将它们编号为$I_1, I_2, \dots, I_k$，在每一个$I_i$中任取一点$\xi_i(i = 1, 2, \dots, k)$，作积分和（也称Riemann和）
|
||
\[\sum_{i = 1}^k f(\xi_i) \sigma(I_i)\]
|
||
记
|
||
\[\norm{\pi} = \max\{\diam (I_1), \dots, \diam(I_k)\}\]
|
||
这里$\diam(I_i)$是矩形$I_i$对角线的长度，称$\norm{\pi}$为分割$\pi$的宽度；$\sigma(I_i)$表示矩形$I_i$的面积。
|
||
|
||
如果存在数$A$使得对任意给定的$\varepsilon > 0$，有$\delta > 0$，凡是$\norm{\pi} < \delta$时，不论值点$\xi_i$在子矩形$I_i$中如何选择，都有
|
||
\[\abs{\sum_{i = 1}^k f(\xi_i) \sigma(I_i) - A} < \varepsilon\]
|
||
则称函数$f$在矩形$I$上可积，并将$A$写作
|
||
\[\iint \limits_I f(x, y) \dif x \dif y \qquad \text{或者} \qquad \int \limits_I f \dif \sigma\]
|
||
称之为$f$在矩形$I$上的二重积分，或简称$f$在$I$上的积分。这里$f$称为被积函数，$I$称为积分区域。
|
||
\end{definition}
|
||
|
||
\begin{theorem}[可积函数的有界性]
|
||
如果$f$在$I$上可积，那么$f$必在$I$上有界。
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[可积函数线性性质]
|
||
设$f, g \in R(D)$，$\alpha, \beta \in \realnum$，则$\alpha f + \beta g \in R(D)$，且
|
||
\[\int \limits_D (\alpha f + \beta g) \dif \sigma = \alpha \int \limits_D f \dif \sigma + \beta \int \limits_D g \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
若$f$和$g$在$I$上可积且$f \geq g$，那么
|
||
\[\int \limits_I f \dif \sigma \geq \int \limits_I g \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
设$D = D_1 \cup D_2$，且$D_1$与$D_2$没有公共内点，若$f(x, y) \in R(D)$，则$f(x, y) \in R(D_1), f(x, y) \in R(D_2)$，且
|
||
\[\iint \limits_D f(x, y) \dif \sigma = \iint \limits_{D_1} f(x, y) \dif \sigma + \iint \limits_{D_2} f(x, y) \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
若$f(x, y) \in R(D)$，则$\abs{f(x, y)} \in R(D)$，并且有
|
||
\[\abs{\iint \limits_D f(x, y) \dif \sigma} \leq \iint \limits_D \abs{f(x, y)} \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
若$f(x, y) \in R(D)$，则$f$有界，设$m \leq f(x, y) \leq M$，则有
|
||
\[m \sigma(D) \leq \iint \limits_D f(x, y) \dif \sigma \leq M \sigma(D)\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[积分中值定理]
|
||
$D \subset \realnum^2$连通、有界闭，$\partial D$为零面积集，$f \in C(D)$，则存在$(\xi, \eta) \in D$，满足
|
||
\[\iint \limits_D f(x, y) \dif \sigma = f(\xi, \eta) \sigma(D)\]
|
||
其中
|
||
\[f(\xi, \eta) = \frac{1}{\sigma(D)} \iint \limits_D f(x, y) \dif \sigma\]
|
||
称为$f(x, y)$在$D$上的平均值。
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
$D \subset \realnum^2$为连通有界闭集，$\partial D$为零面积集，$g$在$D$上不变号，$f, g \in C(D)$。则存在$(\xi, \eta) \in D$，满足
|
||
\[\iint \limits_D f(x, y) g(x, y) \dif \sigma = f(\xi, \eta) \iint \limits_D g(x, y) \dif \sigma\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}
|
||
设$f \in R(D)$，$D$关于$OX$轴对称，则
|
||
\begin{itemize}
|
||
\item 若$f(x, y)$关于$y$为己函数，则$\displaystyle\iint \limits_D f(x, y) \dif \sigma = 0$;
|
||
\item 若$f(x, y)$关于$y$为偶函数，记$D_1$为$D$位于$OX$轴上方的部分，则$\displaystyle\iint \limits_D f(x, y) \dif \sigma = 2 \iint \limits_{D_1} f(x, y) \dif \sigma$。
|
||
\end{itemize}
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[轮换不变性]
|
||
若$D \subset \realnum^2$关于$x, y$是轮换对称的，即$(x, y) \in D \Leftrightarrow (y, x) \in D$，则
|
||
\[\iint \limits_D f(x, y) \dif \sigma = \iint \limits_D f(y, x) \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\section{函数可积性}
|
||
考虑$D = [a, b] \times [c, d]$，$f: D \to \realnum$有界。
|
||
|
||
引入Riemann和记号：令$T$是$D$的一个矩形分割，$\xi = \{\xi_i\}$表示关于分割$T$的小矩形组中任取的一组点集，记
|
||
\[S(T, \xi) = \sum_i f(\xi_i) \sigma (D_i)\eqper\]
|
||
|
||
再引入Darboux和记号，定义上/下和
|
||
\[\overline{S}(T) = \sum_i M_i \sigma (D_i), \underline{S}(T) = \sum_i m_i \sigma(D_i)\]
|
||
其中$M_i = \sup f(D_i), m_i = \inf f(D_i), i = 1, 2, \dots, k$。
|
||
|
||
\begin{lemma}
|
||
对任意的$T$，
|
||
\[\underline{S}(T) \leq S(T, \xi) \leq \overline{S}(T)\eqper\]
|
||
\end{lemma}
|
||
|
||
\begin{lemma}
|
||
设$T^\prime$为分割$T$的加细，则
|
||
\[\underline{S}(T) \leq \underline{S}(T^\prime) \leq \overline{S}(T^\prime) \leq \overline{S}(T)\eqper\]
|
||
\end{lemma}
|
||
|
||
\begin{lemma}
|
||
设$T_1$和$T_2$是$D$的任意两个矩形分割，则
|
||
\[\underline{S}(T_1) \leq \overline{S}(T_2)\eqper\]
|
||
\end{lemma}
|
||
|
||
这说明上下和都是有界的，因此都有上下确界。因此我们可以定义Darboux上下积分：
|
||
\[\underline{\int \limits_{D}} f \dif \sigma = \sup \limits_T \underline{S}(T), \overline{\int \limits_D} f \dif \sigma = \inf \limits_T \overline{S}(T)\eqper\]
|
||
|
||
\begin{lemma}
|
||
设$T_1$和$T_2$是$D$的任意两个矩形分割，则
|
||
\[\underline{S}(T_1) \leq \underline{\int \limits_D} f \dif \sigma \leq \overline{\int \limits_D} f \dif \sigma \leq \overline{S}(T_2)\eqper\]
|
||
\end{lemma}
|
||
|
||
\begin{theorem}[Darboux定理]
|
||
等式
|
||
\[\underline{\int \limits_D} f \dif \sigma = \overline{\int \limits_D} f \dif \sigma\]
|
||
成立当且仅当对任意的$\varepsilon > 0$，都存在分割$T$使得$0 \leq \overline{S}(T) - \underline{S}(T) < \varepsilon$。
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[Riemann定理]
|
||
$f \in R(D)$的充分必要条件是：对任意$\varepsilon > 0$，存在$\delta > 0$，对于任意分割$T$，只要满足$\norm{T} < \delta$，就有$0 \leq \overline{S}(T) - \underline{S}(T) < \varepsilon$。
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[Riemann-Darboux定理]
|
||
$f \in R(D)$的充分必要条件是等式
|
||
\[\underline{\int \limits_D} f \dif \sigma = \overline{\int \limits_D} f \dif \sigma\]
|
||
成立，这时
|
||
\[\int \limits_D f \dif \sigma = \underline{\int \limits_D} f \dif \sigma = \overline{\int \limits_D} f \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\section{Riemann可积函数类}
|
||
\begin{theorem}
|
||
$C(D) \subset R(D)$。
|
||
\end{theorem}
|
||
|
||
\begin{definition}[二维零测集]
|
||
若$E \subset \realnum^2$满足：对任意的$\varepsilon > 0$，存在一列闭矩形$\{D_i\}_{i = 1}^\infty$，使得$E \subset \bigcup \limits_{i = 1}^\infty D_i$且$\sum_{i = 1}^\infty \sigma(D_i) < \varepsilon$，则称$E$为二维零测集。
|
||
\end{definition}
|
||
|
||
\begin{definition}[零面积集]
|
||
若$E \subset \realnum^2$满足：对任意的$\varepsilon > 0$，存在有限个闭矩形$\{D_i\}_{i = 1}^m$，使得$E \subset \bigcup \limits_{i = 1}^m D_i$且$\sum_{i = 1}^m \sigma(D_i) < \varepsilon$，则称$E$为二维零面积集。
|
||
\end{definition}
|
||
|
||
\begin{corollary}
|
||
零面积集是二维零测集。
|
||
\end{corollary}
|
||
|
||
二维零测集的实例：
|
||
\begin{itemize}
|
||
\item 至多可数的点集；
|
||
\item 二维零测集的子集；
|
||
\item 至多可数个二维零测集的并集。
|
||
\end{itemize}
|
||
|
||
零面积集的实例：
|
||
\begin{itemize}
|
||
\item 有限点集；
|
||
\item 零面积集的子集；
|
||
\item 有限多个零面积集的并集。
|
||
\end{itemize}
|
||
|
||
推广：
|
||
\begin{itemize}
|
||
\item 平面上任意有限长直线段是零面积集；
|
||
\item 平面上任意有限长折线是零面积集；
|
||
\item 平面上任意有限长光滑曲线是二维零测集。
|
||
\end{itemize}
|
||
|
||
仿照一元函数的情况，定义
|
||
\begin{definition}[间断点]
|
||
\[D_{is} (f) = \{(x, y) \in D \mid \text{$f$在$(x, y)$点不连续}\}\]
|
||
称为$f$的间断点集。
|
||
\end{definition}
|
||
|
||
\begin{theorem}[Lebesgue定理]
|
||
设$D = [a, b] \times [c, d]$，$f: D \to \realnum$有界，则$f$在$D$上Riemann可积当且仅当$f$的间断点集是二维零测集。
|
||
\end{theorem}
|
||
|
||
\begin{corollary}
|
||
若$f$有界且间断点集是零面积集，则$f$是Riemann可积的。
|
||
\end{corollary}
|
||
|
||
\begin{corollary}
|
||
记$D_0 = \{(x, y) \in D \mid f(x, y) \neq 0\}$。设$f$有界且$D_0$是零面积集，则$f$可积，且
|
||
\[\int \limits_D f \dif \sigma = 0 \eqper\]
|
||
\end{corollary}
|
||
|
||
\begin{corollary}
|
||
设$f, g$都在$D$上有界且仅在零面积集上不想等。若$f \in R(D)$，则必有$g \in R(D)$，且
|
||
\[\int \limits_D f \dif \sigma = \int \limits_D g \dif \sigma \eqper\]
|
||
\end{corollary}
|
||
|
||
\begin{corollary}
|
||
\begin{enumerate}
|
||
\item 若$f \in R(D)$，则$\abs{f} \in R(D)$；
|
||
\item 若$f, g \in R(D)$，则$fg \in R(D)$；
|
||
\item 若$f, g \in R(D)$且$\dfrac{1}{g}$有界，则$\dfrac{f}{g} \in R(D)$。
|
||
\end{enumerate}
|
||
\end{corollary}
|
||
|
||
\section{矩形区域上积分的计算}
|
||
令$D = [a, b] \times [c, d]$，$f \in C(D)$记$I = \dint \limits_D f \dif \sigma$。
|
||
|
||
\begin{enumerate}
|
||
\item 固定$x \in [a, b]$，作为$y$的函数$f(x, y) \in C[c, d]$。令
|
||
\[u(x) = \int_c^d f(x, y) \dif y\]
|
||
易证$u(x) \in C[a,b]$记
|
||
\[A = \int_a^b u(x) \dif x\]
|
||
\item 类似地，固定$y \in [c, d]$可以得到$v(y)$，记
|
||
\[B = \int_c^d v(y) \dif y\]
|
||
\end{enumerate}
|
||
|
||
\begin{theorem}[Fubini定理]
|
||
设$f \in C(D)$，则$I = A = B$：
|
||
\[\int \limits_D f \dif \sigma = \int_a^b \dif x \int_c^d f(x, y) \dif y = \int_c^d \dif y \int_a^b f(x, y) \dif x\]
|
||
其中规定记号
|
||
\begin{align*}
|
||
\int_a^b \dif x \int_c^d f(x, y) \dif y & = \int_a^b \left[\int_c^d f(x, y) \dif y \right]\dif x\\
|
||
\int_c^d \dif y \int_a^b f(x, y) \dif x & = \int_c^d \left[\int_a^b f(x, y) \dif x\right] \dif y\eqper
|
||
\end{align*}
|
||
\end{theorem}
|
||
|
||
\begin{corollary}
|
||
设$f(x, y) = f_1(x) f_2(y)$，其中$f_1 \in C[a, b], f_2 \in C[c, d]$。令$D = [a, b] \times [c, d]$，则$f \in C(D)$，且
|
||
\[\int \limits_D f \dif \sigma = \left[\int_a^b f_1(x) \dif x\right]\left[\int_c^d f_2(y) \dif y\right]\eqper\]
|
||
\end{corollary}
|
||
|
||
\section{有界集合上的二重积分}
|
||
\begin{definition}
|
||
设$D \subset \realnum^2$，函数$f: B \to \realnum$。令
|
||
\[f_D(x, y) = \left\{\begin{aligned}
|
||
& f(x, y), \qquad (x, y) \in D\\
|
||
& 0, \qquad (x, y) \not \in D
|
||
\end{aligned}\right.\]
|
||
\end{definition}
|
||
|
||
\begin{definition}
|
||
任取有界的闭矩形$I \supset D$，如果函数$f_D$在$I$上可积，则说函数$f$在$B$上可积，并把数值$\dint \limits_I f_D \dif \sigma$称为函数$f$在$B$上的（二重）积分，记作
|
||
\[\iint \limits_B f(x, y) \dif x \dif y \quad \text{或者} \quad \int \limits_B f \dif \sigma\]
|
||
\end{definition}
|
||
|
||
在矩形区域上的二重积分性质在一般的有界集合的二重积分上也成立。
|
||
|
||
\begin{theorem}[区域可加性]
|
||
设$D_1, D_2 \subset \realnum^2$都是有界区域，且$D_1 \cap D_2$是零面积集。则$f \in R(D_1 \cup D_2)$当且仅当$f \in R(D_1)$且$f \in R(D_2)$。这时有
|
||
\[\int \limits_{D_1 \cup D_2} f \dif \sigma = \int \limits_{D_1} f \dif \sigma + \int \limits_{D_2} f \dif \sigma\eqper\]
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[一般有界区域上连续函数的可积性]
|
||
设$D \subset \realnum^2$是有界区域，$\partial D$是零面积集，则$C(D) \subset R(D)$。
|
||
\end{theorem}
|
||
|
||
\begin{definition}[一般有界区域的面积]
|
||
设$D \subset \realnum^2$是有界区域，且$\partial D$是零面积集，则$1 \in C(D) \subset R(D)$，可以定义$D$的面积
|
||
\[\sigma(D) = \int \limits_D 1 \dif \sigma\eqper\]
|
||
\end{definition}
|
||
|
||
\begin{proposition}[平面区域面积可加性]
|
||
设$D_1, D_2 \subset \realnum^2$都是有界区域，边界都是零面积集，又设$D_1 \cap D_2$也是零面积集，则
|
||
\[\sigma(D_1 \cup D_2) = \sigma(D_1) + \sigma(D_2) \eqper\]
|
||
\end{proposition}
|
||
|
||
\section{有界集合上积分的计算}
|
||
\begin{theorem}
|
||
设点集$D = \{(x, y) \mid y_1(x) \leq y \leq y_2(x), a \leq x \leq b\}$，其中$y_1, y_2 \in C[a, b]$且$y_1 \leq y_2$，则
|
||
\[\int \limits_D f \dif \sigma = \int_a^b \dif x \int_{y_1(x)}^{y_2(x)} f(x, y) \dif y\eqper\]
|
||
|
||
类似地，如果有$D = \{(x, y) \mid x_1(y) \leq x \leq x_2(y), c \leq y \leq d\}$，其中$x_1, x_2 \in C[c, d]$，且$x_1 \leq x_2$则
|
||
\[\int \limits_D f \dif \sigma = \int_c^d \dif y \int_{x_1(y)}^{x_2(y)} f(x, y) \dif x\eqper\]
|
||
\end{theorem}
|
||
|
||
\section{二重积分换元}
|
||
\begin{theorem}
|
||
设有界闭区域$D \subset \realnum^2$，连续函数$f: D \to \realnum$。$\Omega$是$uOv$平面上的有界闭区域。映射
|
||
\[\bvec{\varphi}: \begin{cases}
|
||
x = x(u, v)\\
|
||
y = y(u, v)
|
||
\end{cases}\]
|
||
是$\Omega$到$D$的一一对应的连续可微映射，满足
|
||
\[\frac{\partial(x, y)}{\partial (u, v)} = \begin{vmatrix}
|
||
\dfrac{\partial x}{\partial u} & \dfrac{\partial x}{\partial v}\\[1em]
|
||
\dfrac{\partial y}{\partial u} & \dfrac{\partial y}{\partial v}
|
||
\end{vmatrix} \neq 0\]
|
||
有换元公式
|
||
\[\iint \limits_D f(x, y) \dif x \dif y = \iint \limits_\Omega f (x(u, v), y(u, v)) \abs{\frac{\partial (x, y)}{\partial (u, v)}} \dif u \dif v\eqper\]
|
||
\end{theorem}
|
||
|
||
下面介绍两种常用的换元：极坐标换元与正交变换。
|
||
|
||
首先介绍极坐标换元。令
|
||
\[\bvec{\varphi}: \begin{cases}
|
||
x = r \cos \theta\\
|
||
y = r \sin \theta
|
||
\end{cases}\]
|
||
这时
|
||
\[\frac{\partial (x, y)}{\partial (r, \theta)} = \begin{vmatrix}
|
||
\cos \theta & -r \sin \theta\\
|
||
\sin \theta & r \cos \theta
|
||
\end{vmatrix}
|
||
= r\]
|
||
如果当$(r, \theta) \in \Omega$时映射$\bvec{\varphi}$将$\Omega$一对一地变为$D$，那么
|
||
\[\iint \limits_D f(x, y) \dif x \dif y = \iint \limits_\Omega f(r\cos \theta, r \sin \theta) r \dif r \dif \theta\eqper\]
|
||
|
||
\begin{example}
|
||
求椭球面
|
||
\(\dfrac{x^2}{a^1} + \dfrac{y^2}{b^2} + \dfrac{z^2}{c^2} = 1\)
|
||
包围的体积$V$。
|
||
\end{example}
|
||
|
||
\begin{proof}[解]
|
||
由对称性，$V$等于$z = 0$平面上部分体积的两倍。在$z = 0$平面上的部分可以看作曲顶柱体，顶部曲面方程
|
||
\[z = c \sqrt{1 - \frac{x^2}{a^2} - \frac{y^2}{b^2}}, D: \frac{x^2}{a^2} + \frac{y^2}{b^2} \leq 1\]
|
||
因此
|
||
\[V = 2 \int \limits_D z \dif \sigma\eqper\]
|
||
|
||
进一步引入广义极坐标变换$x = a r \cos \theta, y = br \sin \theta$。则上面区域$D$转化为矩形区域$\tilde{D}: 0 \leq r \leq 1, 0 \leq \theta \leq 2\pi$。计算Jacobian行列式得$\dif x \dif y = abr \dif r \dif \theta$。因此
|
||
\[V = 2abc \iint \limits_{\tilde{D}} \sqrt{1 - r^2} r \dif r \dif \theta = 4\pi abc \int_0^1 r \sqrt{1 - r^2} \dif r = \frac{4}{3} \pi abc\eqper \qedhere\]
|
||
\end{proof}
|
||
|
||
现在再介绍正交变换。令
|
||
\[\left\{\begin{aligned}
|
||
& u = \frac{ax + by}{\sqrt{a^2 + b^2}}\\
|
||
& v = \frac{ay - bx}{\sqrt{a^2 + b^2}}
|
||
\end{aligned}\right.\]
|
||
那么
|
||
\[\frac{\partial(u, v)}{\partial(x, y)} = \frac{1}{a^2 + b^2} \begin{vmatrix}
|
||
a & b\\
|
||
-b & a
|
||
\end{vmatrix} = 1\]
|
||
注意到正交变换有性质
|
||
\[u^2 + v^2 = \frac{(ax + by)^2 + (bx - ay)^2}{a^2 + b^2} = x^2 + y^2\]
|
||
这在变换积分域时通常能给出不少好处。
|
||
|
||
\begin{example}
|
||
设$f \in C(\realnum)$，$D = \{(x, y) \mid x^2 + y^2 \leq R^2\}$。将二重积分
|
||
\[\iint \limits_{x^2 + y^2 \leq R^2} f(ax + by + c) \dif x \dif y\]
|
||
约化为一重积分。
|
||
\end{example}
|
||
|
||
\begin{proof}[解]
|
||
引入正交变换
|
||
\[\left\{\begin{aligned}
|
||
& u = \frac{ax + by}{\sqrt{a^2 + b^2}}\\
|
||
& v = \frac{ay - bx}{\sqrt{a^2 + b^2}}
|
||
\end{aligned}\right.\]
|
||
因此
|
||
\begin{align*}
|
||
\text{原式} & = \iint \limits_{u^2 + v^2 \leq R^2} f\left(\sqrt{a^2 + b^2} u + c\right) \dif u \dif v\\
|
||
& = \int_{-R}^R \dif u \int_{-\sqrt{R^2 - u^2}}^{\sqrt{R^2 - u^2}} f \left(\sqrt{a^2 + b^2}u + c\right) \dif v\\
|
||
& = 2\int_{-R}^R \sqrt{R^2 - u^2} f\left(\sqrt{a^2 + b^2}u + c\right) \dif u\eqper \qedhere
|
||
\end{align*}
|
||
\end{proof}
|
||
|
||
\section{三重积分}
|
||
\begin{definition}[三重积分]
|
||
令$\Omega = [a, b] \times [c, d] \times [g, h]$是$\realnum^3$中的长方体。$f: \Omega \to \realnum$。将$\Omega$做有限的分割$T = \pi_x \times \pi_y \times \pi_z$，其中
|
||
\begin{align*}
|
||
& \pi_x: a = x_0 < x_1 < \dots < x_n = b\\
|
||
& \pi_y: c = y_0 < y_1 < \dots < y_n = d\\
|
||
& \pi_z: g = z_0 < z_1 < \dots < z_k = h
|
||
\end{align*}
|
||
将$\Omega$分割为$nmk$个小长方体。将他们重新编号为$\Omega_1, \Omega_2, \dots, \Omega_{nmk}$。在每个$\Omega_i$中任取一点$\xi_i$，作Riemann和
|
||
\[\sum_{i = 1}^{nmk} f(\xi_i) \mu(\Omega_i)\]
|
||
记
|
||
\[\norm{T} = \max\{\diam (\Omega_1), \dots, \diam (\Omega_{nmk})\}\]
|
||
这里$\diam (\Omega_i)$是长方体$\Omega_i$对角线的长度。称$\norm{T}$为分割$T$的直径；$\mu(\Omega_i)$表示长方体$\Omega_i$的体积。
|
||
|
||
如果存在数$A$使得对任意给定的$\varepsilon > 0$，有$\delta > 0$，凡是$\norm{T} < \delta$时，不论值点$\xi_i$在子长方体$\Omega_i$中如何选择，都有
|
||
\[\abs{\sum_{i = 1}^{nmk} f(\xi_i) \mu(\Omega_i) - A} < \varepsilon\]
|
||
则称$A$为$f$在$\Omega$上的积分，记作
|
||
\[\iiint \limits_\Omega f(x, y, z) \dif x \dif y \dif z \qquad \text{或者} \qquad \int \limits_\Omega f \dif \mu\eqper\]
|
||
\end{definition}
|
||
|
||
类似二维的情况，我们可以定义零测集与零体积集：
|
||
\begin{definition}[三维零测集]
|
||
若$E \subset \realnum^3$满足：对任意的$\varepsilon > 0$，存在一系列闭长方体$\{\Omega_i\}_{i = 1}^\infty$使得$E \subset \bigcup \limits_{i = 1}^\infty \Omega_i$且$\sum_{i = 1}^\infty \mu(\Omega_i) < \varepsilon$，则称$E$为三维零测集。
|
||
\end{definition}
|
||
|
||
\begin{definition}[零体积集]
|
||
若$E \subset \realnum^3$满足：对任意的$\varepsilon > 0$，存在有限的闭长方体$\{\Omega_i\}_{i = 1}^m$使得$E \subset \bigcup \limits_{i = 1}^m \Omega_i$且$\sum_{i = 1}^m \mu(\Omega_i) < \varepsilon$，则称$E$为零体积集。
|
||
\end{definition}
|
||
|
||
\begin{theorem}[Lebesgue定理]
|
||
对于长方体$\Omega$上的有界函数$f$，积分
|
||
\(\dint \limits_\Omega f \dif \mu\)
|
||
存在的充分必要条件是$f$在$\Omega$上的间断点集是一零测集。
|
||
\end{theorem}
|
||
|
||
\begin{theorem}[Fubini定理]
|
||
设$f \in C(\Omega)$，$\Omega = [a, b] \times [c, d] \times [g, h]$，则
|
||
\begin{align*}
|
||
\int \limits_\Omega f \dif \mu & = \int_a^b \dif x \int_c^d \dif y \int_g^h f(x, y, z) \dif z\\
|
||
& = \int_c^d \dif y \int_a^b \dif x \int_g^h f(x, y, z) \dif z\\
|
||
& = \int_g^h \dif z \int_a^b \dif x \int_c^d f(x, y, z) \dif y\\
|
||
& = \int_a^b \dif x \int_g^h \dif z \int_c^d f(x, y, z) \dif y\\
|
||
& = \int_c^d \dif y \int_g^h \dif z \int_a^b f(x, y, z) \dif x\\
|
||
& = \int_g^h \dif z \int_c^d \dif y \int_a^b f(x, y, z) \dif x\eqper1
|
||
\end{align*}
|
||
\end{theorem}
|
||
|
||
一般有界区域上的三重积分和及三重积分的性质与二重积分完全类似。
|
||
|
||
对一般有界区域上的三重积分，设$\Omega \subset \realnum^3$是有界区域，$f: \Omega \to \realnum$。引入延拓函数
|
||
\[f_\Omega (x, y, z) = \begin{cases}
|
||
f(x, y, z), \quad (x, y, z) \in \Omega\\
|
||
0, \quad (x, y, z) \not \in \Omega
|
||
\end{cases}\]
|
||
记$\Omega_M = [-M, M] \times [-M, M] \times [-M, M] \supset \Omega$。如果$f_\Omega \in R(\Omega_M)$，则称$f \in R(\Omega)$。定义$f$在$\Omega$上的积分值
|
||
\[\int_\Omega f \dif \mu = \int \limits_{\Omega_M} f_\Omega \dif \mu\]
|
||
再记$D_M = [-M, M] \times [-M, M]$，结合Fubini定理可得
|
||
\begin{align*}
|
||
\int \limits_\Omega f \dif \mu & = \iint \limits_{D_M} \dif x \dif y \int_{-M}^M f_\Omega(x, y, z) \dif z\\
|
||
& = \int_{-M}^M \dif z \iint \limits_{D_M} f_\Omega(x, y, z) \dif x \dif y\eqper
|
||
\end{align*}
|
||
|
||
下面我们考虑在一般区域上的三重积分的计算。
|
||
\begin{theorem}
|
||
设有界集$\Omega \subset \realnum^3$有体积，有界的函数$f: \Omega \to \realnum$连续。
|
||
|
||
\begin{enumerate}
|
||
\item 设$\Omega$在$xy$平面上的垂直投影为$D$，且当$(x, y) \in D$时，过这一点且垂直于$D$的直线与$V$交成一个区间$[z_1(x, y), z_2(x, y)]$，即$\Omega$可以表示为
|
||
\[\{(x, y, z) \mid z_1(x, y) \leq z \leq z_2(x, y), (x, y) \in D\}\]
|
||
那么
|
||
\[\int \limits_\Omega f \dif \mu = \iint \limits_D \dif x \dif y \int_{z_1(x, y)}^{z_2(x, y)} f(x, y, z) \dif z\text{；}\]
|
||
\item 设$\Omega$在$z$轴上的垂直投影为区间$J$，且当$z \in J$时，通过点$(0, 0, z)$同时垂直于$z$轴的平面与$\Omega$交成的图形在$xy$平面上的垂直投影是一个有面积的点集$D$，即$\Omega$可以表示为
|
||
\[\{(x, y, z) \mid (x, y) \in \bvec{D}(z), z \in J\}\]
|
||
那么
|
||
\[\int \limits_\Omega f \dif \mu = \int \limits_J \dif z \iint \limits_D f(x, y, z) \dif x \dif y\eqper\]
|
||
\end{enumerate}
|
||
\end{theorem}
|
||
|
||
最后我们仿照二元积分的换元公式，给出三重积分的换元公式。
|
||
\begin{theorem}
|
||
设函数$f: \Omega \to \realnum$连续，映射
|
||
\[\bvec{\varphi}: \begin{cases}
|
||
x = x(u, v, w)\\
|
||
y = y(u, v, w)\\
|
||
z = z(u, v, w)
|
||
\end{cases}, (u, v, w) \in \tilde{\Omega}\]
|
||
是$\tilde{\Omega}$到$\Omega$的一一对应，且$\bvec{\varphi} \in C^1 (\tilde{\Omega}, \realnum^3)$满足
|
||
\[\det J\bvec{\varphi} = \frac{\partial (x, y, z)}{\partial (u, v, w)} = \begin{vmatrix}
|
||
\dfrac{\partial x}{\partial u} & \dfrac{\partial x}{\partial v} & \dfrac{\partial x}{\partial w}\\[1em]
|
||
\dfrac{\partial y}{\partial u} & \dfrac{\partial y}{\partial v} & \dfrac{\partial y}{\partial w}\\[1em]
|
||
\dfrac{\partial z}{\partial u} & \dfrac{\partial z}{\partial v} & \dfrac{\partial z}{\partial w}\\
|
||
\end{vmatrix} \neq 0\]
|
||
则有
|
||
\[\int \limits_\Omega f \dif \mu = \int \limits_{\tilde{\Omega}} f \circ \bvec{\varphi} \abs{\det J \bvec{\varphi}} \dif \mu\]
|
||
也即
|
||
\[\iiint \limits_\Omega f(x, y, z) \dif x \dif y \dif z = \iiint \limits_{\tilde{\Omega}} f(x(u, v, w), y(u, v, w), z(u, v, w)) \abs{\frac{\partial (x, y, z)}{\partial (u, v, w)}} \dif u \dif v \dif w \eqper\]
|
||
\end{theorem}
|
||
|
||
下面给出这换元公式的几个常见的特例。
|
||
|
||
第一个特例是柱坐标换元。设映射
|
||
\[\bvec{\varphi}: \begin{cases}
|
||
x = r \cos \theta\\
|
||
y = r \sin \theta\\
|
||
z = z
|
||
\end{cases}\]
|
||
这时
|
||
\[\frac{\partial (x, y, z)}{\partial (r, \theta, z)} \begin{vmatrix}
|
||
\cos \theta & \sin \theta & 0\\
|
||
-r \sin \theta & r \cos \theta & 0\\
|
||
0 & 0 & 1
|
||
\end{vmatrix}= r\]
|
||
则
|
||
\[\iiint \limits_\Omega f \dif \mu = \iiint \limits_{\tilde{\Omega}} f(r \cos \theta, r \sin \theta, z) r \dif r \dif \theta \dif z\eqper\]
|
||
|
||
第二个特例是球坐标变换。设映射
|
||
\[\bvec{\varphi}: \begin{cases}
|
||
x = r \sin \theta \cos \varphi\\
|
||
y = r \sin \theta \sin \varphi\\
|
||
z = r \cos \theta
|
||
\end{cases}\]
|
||
这时
|
||
\[\frac{\partial (x, y, z)}{\partial (r, \theta, z)} = r^2 \sin \theta\]
|
||
则
|
||
\[\int \limits_\Omega f \dif \mu = \iiint \limits_{\tilde{\Omega}} f(r \sin \theta \cos \varphi, r \sin \theta \sin \varphi, r \cos \theta)\eqper\] |