连续性

定义

\(f:D\subset \mathbb R\to \mathbb R,\ a\in D\):

\[f\ \text{在}\ a\ \text{连续}\ \iff\ \forall\varepsilon>0\ \exists\delta>0\ \forall x:\ \bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\Rightarrow |f(x)-f(a)|<\varepsilon \]

三种等价刻画(明确列出三者等价):

  1. 极限式:\(\displaystyle \lim_{x\to a} f(x)=f(a)\)
  2. \(\varepsilon\)-\(\delta\) 式:\(\forall\varepsilon>0\ \exists\delta>0\ \forall x:\ \bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\Rightarrow |f(x)-f(a)|<\varepsilon\)
  3. 序列式:\(\forall (x_n)_{n\in\mathbb N}\subseteq D:\ \bigl(x_n\to a\bigr)\Rightarrow\bigl(f(x_n)\to f(a)\bigr)\)

\[1\ \iff\ 2\ \iff\ 3 \]

其中序列式展开为

\[\forall (x_n)_{n\in\mathbb N}\subseteq D:\ \Bigl(\forall\varepsilon>0\ \exists N\in\mathbb N\ \forall n\ge N:\ |x_n-a|<\varepsilon\Bigr)\Rightarrow\Bigl(\forall\varepsilon>0\ \exists N\in\mathbb N\ \forall n\ge N:\ |f(x_n)-f(a)|<\varepsilon\Bigr) \]

否定式:

\[\neg\bigl(f\ \text{在}\ a\ \text{连续}\bigr)\ \iff\ \exists\varepsilon>0\ \forall\delta>0\ \exists x:\ \bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\ \wedge\ \neg\bigl(|f(x)-f(a)|<\varepsilon\bigr) \]

邻域语言:

\[\forall\varepsilon>0\ \exists\delta>0:\ f\bigl(U_\delta(a)\cap D\bigr)\subset U_\varepsilon\bigl(f(a)\bigr) \]

全局连续
设 \(f:D\subset\mathbb R\to\mathbb R\):

\[f\ \text{在}\ D\ \text{上连续}\ :\iff\ \forall a\in D:\ f\ \text{在}\ a\ \text{连续} \]

展开即

\[\forall a\in D\ \forall\varepsilon>0\ \exists\delta>0\ \forall x:\ \bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\Rightarrow|f(x)-f(a)|<\varepsilon \]

在 \(D\) 的每一点连续 \(\implies\) 在 \(D\) 上连续:

\[\bigl(\forall a\in D:\ f\ \text{在}\ a\ \text{连续}\bigr)\Rightarrow\bigl(f\ \text{在}\ D\ \text{上连续}\bigr) \]

(实为 \(\iff\),因右端本就定义为左端。)

若存在一个不连续点

\[\bigl(\exists a\in D:\ \neg(f\ \text{在}\ a\ \text{连续})\bigr)\Rightarrow\neg\bigl(f\ \text{在}\ D\ \text{上连续}\bigr) \]

等价的逆否/取否形式:

\[f\ \text{在}\ D\ \text{上连续}\Rightarrow\neg\bigl(\exists a\in D:\ \neg(f\ \text{在}\ a\ \text{连续})\bigr)\iff\forall a\in D:\ f\ \text{在}\ a\ \text{连续} \]

三个入门例题

  1. 常数函数 \(f(x)=c\):平凡.

\[|f(x)-f(a)|=|c-c|=0<\varepsilon\qquad(\forall x\in D,\ \forall\delta>0) \]

即 \(\forall\varepsilon>0\ \exists\delta>0\)(可任取,如 \(\delta:=1\))\(\forall x:\ \bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\Rightarrow 0<\varepsilon\).

  1. 线性函数 \(f(t)=mt+b\):取

\[\delta=\frac{\varepsilon}{1+|m|}\ (>0) \]

验证:

\[|f(t)-f(a)|=|mt+b-(ma+b)|=|m|\cdot|t-a|<|m|\delta=\frac{|m|}{1+|m|}\varepsilon<\varepsilon \]

\[\delta=\delta(\varepsilon)\quad\text{即}\quad\forall\varepsilon>0\ \exists\delta>0\ \forall a\in D\ \forall t\in D:\ \bigl(|t-a|<\delta\bigr)\Rightarrow\bigl(|f(t)-f(a)|<\varepsilon\bigr) \]

  1. 抛物线 \(f(t)=t^2\):由

\[|t^2-a^2|=|t-a||t+a| \]

先限制 \(|t-a|<1\) 得 \(|t+a|=|(t-a)+2a|\le|t-a|+2|a|<1+2|a|\),取

\[\delta=\frac{\varepsilon}{1+2|a|}\quad(\text{可设 }\varepsilon<1) \]

(严格写法应同时兜住 \(|t-a|<1\):\(\displaystyle\delta=\min\Bigl\{1,\ \frac{\varepsilon}{1+2|a|}\Bigr\}\),即 \(\delta\le1\ \wedge\ \delta\le\frac{\varepsilon}{1+2|a|}\)。)

验证:

\[|t-a|<\delta\ \Rightarrow\ |f(t)-f(a)|=|t-a||t+a|<\delta\,(1+2|a|)\le\varepsilon \]

而

\[\delta=\delta(\varepsilon,a),\qquad\neg\Bigl(\forall\varepsilon>0\ \exists\delta>0\ \forall a\in D\ \forall t\in D:\ |t-a|<\delta\Rightarrow|f(t)-f(a)|<\varepsilon\Bigr) \]

连续性的代数定理:
(序列式):\(f,g\) 在 \(a\) 连续,即

\[\forall(x_n)\subseteq D:\ \bigl(x_n\to a\bigr)\Rightarrow\bigl(f(x_n)\to f(a)\bigr),\qquad \forall(x_n)\subseteq D:\ \bigl(x_n\to a\bigr)\Rightarrow\bigl(g(x_n)\to g(a)\bigr) \]

  • 四则运算:

\[\bigl(f\ \text{在}\ a\ \text{连续}\ \wedge\ g\ \text{在}\ a\ \text{连续}\bigr)\Rightarrow\bigl(f\pm g\ \text{在}\ a\ \text{连续}\ \wedge\ fg\ \text{在}\ a\ \text{连续}\bigr) \]

\[\bigl(f\ \text{在}\ a\ \text{连续}\ \wedge\ g\ \text{在}\ a\ \text{连续}\ \wedge\ g(a)\ne0\bigr)\Rightarrow\Bigl(\frac fg\ \text{在}\ a\ \text{连续}\Bigr) \]

序列写法:

\[\lim(f\pm g)(x_n)=f(a)\pm g(a),\quad \lim(fg)(x_n)=f(a)g(a),\quad g(a)\ne0\Rightarrow\lim\frac{f(x_n)}{g(x_n)}=\frac{f(a)}{g(a)} \]

  • 线性组合 \(\lambda f+\mu g\) 连续:

\[\bigl(\lambda,\mu\in\mathbb R\ \wedge\ f\ \text{在}\ a\ \text{连续}\ \wedge\ g\ \text{在}\ a\ \text{连续}\bigr)\Rightarrow\bigl(\lambda f+\mu g\ \text{在}\ a\ \text{连续}\bigr) \]

  • 复合:

\[\bigl(f\ \text{在}\ a\ \text{连续}\ \wedge\ g\ \text{在}\ f(a)\ \text{连续}\bigr)\Rightarrow\bigl(g\circ f\ \text{在}\ a\ \text{连续}\bigr) \]

即 \(\forall(x_n)\subseteq D:\ x_n\to a\Rightarrow f(x_n)\to f(a)\Rightarrow g(f(x_n))\to g(f(a))\)。

  • 绝对值:

\[\big||f(x)|-|f(a)|\big|\le|f(x)-f(a)| \]

(三角不等式)\(\implies |f|\) 连续:

\[\bigl(f\ \text{在}\ a\ \text{连续}\bigr)\Rightarrow\bigl(|f|\ \text{在}\ a\ \text{连续}\bigr) \]

即 \(\forall\varepsilon>0\ \exists\delta>0\ \forall x:\ \bigl(|x-a|<\delta\Rightarrow|f(x)-f(a)|<\varepsilon\bigr)\Rightarrow\bigl(|x-a|<\delta\Rightarrow\big||f(x)|-|f(a)|\big|<\varepsilon\bigr)\)。

  • 推论:\(C(D)\) 是线性空间,甚至是代数:

\[C(D):=\{f:D\to\mathbb R\mid f\ \text{在}\ D\ \text{上连续}\} \]

\[f,g\in C(D)\ \wedge\ \lambda,\mu\in\mathbb R\ \Rightarrow\ \bigl(\lambda f+\mu g\in C(D)\bigr)\ \wedge\ \bigl(fg\in C(D)\bigr) \]

  • 推论:所有多项式连续:

\[p(t)=\sum_{k=0}^{n}c_k t^k\ \Rightarrow\ p\in C(\mathbb R) \]

(由 \(t\mapsto t\) 连续 \(\ \wedge\) 加、数乘、乘封闭 \(\ \Rightarrow\ t^k\) 连续 \(\ \Rightarrow\) 有限和连续)

\[\bigl(\text{有理函数}\ \vee\ \text{三角}\ \vee\ \text{反三角}\ \vee\ \exp\ \vee\ \ln\ \vee\ \text{双曲}\ \vee\ \text{幂函数}\bigr)\Rightarrow\text{在其定义域上连续} \]

反例收尾

  • \(\mathrm{sgn}\) 在 \(0\) 不连续(两条序列):

\[x_n=\frac1n\to0\ \Rightarrow\ \mathrm{sgn}(x_n)=1\to1\ \ne\ 0=\mathrm{sgn}(0) \]

\[x_n'=-\frac1n\to0\ \Rightarrow\ \mathrm{sgn}(x_n')=-1\to-1\ \ne\ 0=\mathrm{sgn}(0) \]

即

\[\neg\bigl(\mathrm{sgn}\ \text{在}\ 0\ \text{连续}\bigr)\iff\exists(x_n)\subseteq D:\ \bigl(x_n\to0\bigr)\ \wedge\ \neg\bigl(\mathrm{sgn}(x_n)\to\mathrm{sgn}(0)\bigr) \]

  • Gauß 括号 \([t]\) 在每个整数点不连续:

\[\left[m-\frac1n\right]=m-1\ne m=[m]\qquad(m\in\mathbb Z) \]

即 \(\forall m\in\mathbb Z\):

\[x_n=m-\frac1n\to m\ \wedge\ [x_n]=m-1\to m-1\ \ne\ m=[m] \]

\[\Rightarrow\ \neg\bigl(t\mapsto[t]\ \text{在}\ m\ \text{连续}\bigr)\iff\exists(x_n):\ \bigl(x_n\to m\bigr)\ \wedge\ \neg\bigl([x_n]\to[m]\bigr) \]

反例库(本篇的看点)

记号约定:\(D\subset\mathbb R\),\(a\in D\)。
连续:\(C(f,a):\iff \forall\varepsilon>0\ \exists\delta>0\ \forall x:\bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\Rightarrow|f(x)-f(a)|<\varepsilon\)
序列式:\(C(f,a)\iff \forall(x_n)\subseteq D:\bigl(x_n\to a\bigr)\Rightarrow\bigl(f(x_n)\to f(a)\bigr)\)
不连续(\(\neg\) 展开,德·摩根:\(\neg\forall=\exists\neg\),\(\neg(P\Rightarrow Q)=P\wedge\neg Q\)):

\[\neg C(f,a)\iff\exists\varepsilon>0\ \forall\delta>0\ \exists x:\bigl(x\in D\ \wedge\ |x-a|<\delta\bigr)\ \wedge\ \neg\bigl(|f(x)-f(a)|<\varepsilon\bigr) \]

\[\neg C(f,a)\iff\exists(x_n)\subseteq D:\bigl(x_n\to a\bigr)\ \wedge\ \neg\bigl(f(x_n)\to f(a)\bigr) \]

  1. Dirichlet 函数 —— 处处不连续

\[\chi_{\mathbb Q}(x)=\begin{cases}1,&x\in\mathbb Q\\0,&x\notin\mathbb Q\end{cases} \]

处处不连续:

\[\forall a\in\mathbb R:\ \neg C(\chi_{\mathbb Q},a) \]

分两种情形(\(\mathbb Q\) 与 \(\mathbb R\setminus\mathbb Q\) 均在 \(\mathbb R\) 中稠密):

\[a\in\mathbb Q:\ \exists(x_n)\subseteq\mathbb R\setminus\mathbb Q:\ \bigl(x_n\to a\bigr)\ \wedge\ \bigl(\chi_{\mathbb Q}(x_n)=0\to0\bigr)\ \wedge\ \neg\bigl(0=1=\chi_{\mathbb Q}(a)\bigr) \]

\[a\notin\mathbb Q:\ \exists(x_n)\subseteq\mathbb Q:\ \bigl(x_n\to a\bigr)\ \wedge\ \bigl(\chi_{\mathbb Q}(x_n)=1\to1\bigr)\ \wedge\ \neg\bigl(1=0=\chi_{\mathbb Q}(a)\bigr) \]

\(\varepsilon\)-\(\delta\) 直证(取 \(\varepsilon=\tfrac12\)):

\[\exists\varepsilon=\tfrac12>0\ \forall\delta>0\ \exists x:\ \bigl(|x-a|<\delta\bigr)\ \wedge\ \bigl|\chi_{\mathbb Q}(x)-\chi_{\mathbb Q}(a)\bigr|=1\ \wedge\ \neg\bigl(1<\tfrac12\bigr) \]

  1. Thomae 函数 —— 在无理点连续、有理点不连续

\[T(x)=\begin{cases}\dfrac1q,&x=\dfrac pq,\ p\in\mathbb Z,\ q\in\mathbb N,\ \gcd(p,q)=1\\[2mm]0,&x\notin\mathbb Q\end{cases}\qquad\bigl(T(0)=1,\ \text{因 }0=\tfrac01\bigr) \]

结论:

\[\bigl(a\notin\mathbb Q\Rightarrow C(T,a)\bigr)\ \wedge\ \bigl(a\in\mathbb Q\Rightarrow\neg C(T,a)\bigr) \]

(3a) 无理点连续。设 \(a\notin\mathbb Q\),\(\varepsilon>0\):

\[S_\varepsilon:=\Bigl\{\tfrac pq\in[a-1,a+1]\ \Big|\ q\le\tfrac1\varepsilon\Bigr\},\qquad S_\varepsilon<\infty\ \wedge\ a\notin S_\varepsilon \]

\[\delta:=\min\Bigl\{1,\ \tfrac12\min_{r\in S_\varepsilon}|a-r|\Bigr\}>0 \]

\[\bigl(|x-a|<\delta\bigr)\Rightarrow\Bigl[\bigl(x\notin\mathbb Q\ \wedge\ |T(x)-T(a)|=0<\varepsilon\bigr)\ \vee\ \bigl(x=\tfrac pq\ \wedge\ q>\tfrac1\varepsilon\ \wedge\ |T(x)-T(a)|=\tfrac1q<\varepsilon\bigr)\Bigr] \]

(关键:\(\neg\bigl(q\le\tfrac1\varepsilon\bigr)\Rightarrow q>\tfrac1\varepsilon\Rightarrow\tfrac1q<\varepsilon\);有限集避让保证 \(\delta\) 存在。)

(3b) 有理点不连续。设 \(a=\tfrac pq\),\(T(a)=\tfrac1q>0\):

\[\exists(x_n)\subseteq\mathbb R\setminus\mathbb Q:\ \bigl(x_n\to a\bigr)\ \wedge\ \bigl(T(x_n)=0\to0\bigr)\ \wedge\ \neg\Bigl(0=\tfrac1q=T(a)\Bigr) \]

\[\Rightarrow\ \neg C(T,a) \]

  1. \(\displaystyle f(x)=x\sin\frac1x\)(补 \(f(0)=0\))—— 在 \(0\) 连续

\[f(x)=\begin{cases}x\sin\dfrac1x,&x\ne0\\[1mm]0,&x=0\end{cases} \]

\[|f(x)-f(0)|=\Bigl|x\sin\tfrac1x\Bigr|=|x|\cdot\Bigl|\sin\tfrac1x\Bigr|\le|x|\cdot1=|x| \]

取 \(\delta=\varepsilon\):

\[\forall\varepsilon>0\ \exists\delta=\varepsilon>0\ \forall x:\bigl(|x-0|<\delta\bigr)\Rightarrow\bigl(|f(x)-f(0)|\le|x|<\varepsilon\bigr) \]

\[\Rightarrow\ C(f,0) \]

且 \(x\ne0\) 处由复合定理连续(\(x\mapsto\tfrac1x\) 在 \(\mathbb R\setminus\{0\}\) 连续 \(\ \wedge\ \sin\) 连续 \(\ \wedge\) 乘积连续):

\[\bigl(\forall a\ne0:\ C(f,a)\bigr)\ \wedge\ C(f,0)\ \Rightarrow\ f\in C(\mathbb R) \]

(注:\(\delta=\varepsilon\) 与 \(a\) 无关;但 \(f\) 在 \(0\) 不可导,与连续性无关。)

  1. \(\displaystyle \sin\frac1x\)(补 \(f(0)=0\))—— 在 \(0\) 本质间断

\[g(x)=\begin{cases}\sin\dfrac1x,&x\ne0\\[1mm]0,&x=0\end{cases} \]

(5a) \(0\) 处不连续(两条序列分别撞向 \(1\) 与 \(-1\)):

\[x_n=\frac1{\frac\pi2+2\pi n}\to0\ \wedge\ g(x_n)=\sin\Bigl(\tfrac\pi2+2\pi n\Bigr)=1\to1\ \ne\ 0=g(0) \]

\[x_n'=\frac1{\frac{3\pi}2+2\pi n}\to0\ \wedge\ g(x_n')=\sin\Bigl(\tfrac{3\pi}2+2\pi n\Bigr)=-1\to-1\ \ne\ 0=g(0) \]

\[\Rightarrow\ \exists(x_n):\bigl(x_n\to0\bigr)\ \wedge\ \neg\bigl(g(x_n)\to g(0)\bigr)\ \Rightarrow\ \neg C(g,0) \]

(5b) 本质间断(第二类间断:极限不存在,且不为 \(\pm\infty\)):

\[\neg\Bigl(\exists L\in\mathbb R:\ \lim_{x\to0}g(x)=L\Bigr)\ \wedge\ \neg\Bigl(\lim_{x\to0}g(x)=+\infty\Bigr)\ \wedge\ \neg\Bigl(\lim_{x\to0}g(x)=-\infty\Bigr) \]

\[\bigl(\forall x\ne0:\ |g(x)|\le1\bigr)\ \wedge\ \Bigl(\limsup_{x\to0}g(x)=1\Bigr)\ \wedge\ \Bigl(\liminf_{x\to0}g(x)=-1\Bigr)\ \wedge\ (1\ne-1) \]

与跳跃

  1. \(\displaystyle g(x)=\sin\frac1x\)(补 \(g(0)=0\))—— 在 \(0\) 本质间断

函数定义(\(D=\mathbb R\)):

\[g(x)=\begin{cases}\sin\dfrac1x,&x\ne0\\[1mm]0,&x=0\end{cases} \]

(5a) \(x\ne0\) 处连续(复合定理,非手算)

\[\Bigl(x\mapsto\tfrac1x\Bigr)\in C(\mathbb R\setminus\{0\})\ \wedge\ \sin\in C(\mathbb R)\ \Rightarrow\ \bigl(\sin(\frac1x)\bigr)\in C(\mathbb R\setminus\{0\}) \]

\[\forall a\in\mathbb R\setminus\{0\}:\ C(g,a) \]

(5b) \(0\) 处不连续 —— 序列式

\[x_n:=\frac1{\frac\pi2+2\pi n}\ \Rightarrow\ \Bigl(x_n>0\Bigr)\ \wedge\ \Bigl(x_n\to0\Bigr)\ \wedge\ \Bigl(g(x_n)=\sin\bigl(\tfrac\pi2+2\pi n\bigr)=1\to1\Bigr) \]

\[x_n':=\frac1{\frac{3\pi}2+2\pi n}\ \Rightarrow\ \Bigl(x_n'>0\Bigr)\ \wedge\ \Bigl(x_n'\to0\Bigr)\ \wedge\ \Bigl(g(x_n')=\sin\bigl(\tfrac{3\pi}2+2\pi n\bigr)=-1\to-1\Bigr) \]

\[\Bigl(x_n\to0\ \wedge\ g(x_n)\to1\Bigr)\ \wedge\ \Bigl(x_n'\to0\ \wedge\ g(x_n')\to-1\Bigr)\ \wedge\ \bigl(g(0)=0\bigr)\ \wedge\ \bigl(1\ne0\bigr) \]

\[\Rightarrow\ \exists(x_n)\subseteq\mathbb R:\bigl(x_n\to0\bigr)\ \wedge\ \neg\bigl(g(x_n)\to g(0)\bigr)\ \Rightarrow\ \neg C(g,0) \]

(5c) \(0\) 处不连续 —— \(\varepsilon\)-\(\delta\) 否定式
取 \(\varepsilon=\tfrac12\)。由 Archimedes 原理:\(\forall\delta>0\ \exists n\in\mathbb N:\ \tfrac\pi2+2\pi n>\tfrac1\delta\),令 \(x=\dfrac1{\frac\pi2+2\pi n}\),则

\[\bigl(|x-0|=x<\delta\bigr)\ \wedge\ \bigl(g(x)=1\bigr)\ \wedge\ \bigl(g(0)=0\bigr)\ \wedge\ \bigl|g(x)-g(0)\bigr|=1\ \wedge\ \neg\bigl(1<\tfrac12\bigr) \]

\[\Rightarrow\ \exists\varepsilon=\tfrac12>0\ \forall\delta>0\ \exists x:\bigl(|x|<\delta\bigr)\ \wedge\ \neg\bigl(|g(x)-g(0)|<\varepsilon\bigr)\ \Rightarrow\ \neg C(g,0) \]

定理

定理 A(Weierstrass)

记(\([a,b]\) 闭有界,\(\mathbb R\) 完备):

\[C(f):\forall t\in[a,b]\ \forall\varepsilon>0\ \exists\delta>0\ \forall x:\bigl(x\in[a,b]\wedge|x-t|<\delta\bigr)\Rightarrow|f(x)-f(t)|<\varepsilon \]

\[B(f):\exists K\ \forall x\in[a,b]:\ |f(x)|\le K,\qquad E(f):\exists x_{\min},x_{\max}\in[a,b]\ \forall x:\ m=f(x_{\min})\le f(x)\le f(x_{\max})=M \]

\[\text{定理}\quad C(f)\Rightarrow B(f)\wedge E(f) \]

  1. 有界性(反证)

\[\neg B(f)\iff\forall K\ \exists x\in[a,b]:\ \neg(|f(x)|\le K)\iff\forall K\ \exists x\in[a,b]:\ |f(x)|>K \]

取 \(K=n\in\mathbb N\):\(\exists(x_n)\subseteq[a,b]\ \forall n:\bigl(|f(x_n)|>n\bigr)\wedge\bigl(a\le x_n\le b\bigr)\)

\[\overset{\mathrm{BW}}{\Rightarrow}\exists(x_{n_k})\ \exists t:\bigl(x_{n_k}\to t\bigr)\wedge\bigl(\forall k:\ a\le x_{n_k}\le b\bigr)\wedge\bigl(n_k\ge k\bigr)\ \overset{[a,b]\ \text{闭}}{\Rightarrow}\ t\in[a,b] \]

\[\bigl(x_{n_k}\to t\bigr)\wedge C(f)\Rightarrow f(x_{n_k})\to f(t)\Rightarrow\exists k_0\ \forall k\ge k_0:\ |f(x_{n_k})|<|f(t)|+1 \]

取 \(k\ge k_0\wedge k>|f(t)|+1\):

\[\underbrace{|f(x_{n_k})|>n_k\ge k>|f(t)|+1}_{\neg B(f)}\ \wedge\ \underbrace{|f(x_{n_k})|<|f(t)|+1}_{\text{收敛}}\ \Rightarrow\ \bot \]

\[\implies\ \neg B(f)\Rightarrow\bot\ \Rightarrow\ B(f) \]

  1. 最值(用结论 1;\(M:=\sup f\))

\[M=\sup f\iff\Bigl(\forall x:\ f(x)\le M\Bigr)\wedge\Bigl(\forall\varepsilon>0\ \exists x:\ f(x)>M-\varepsilon\Bigr) \]

\[\neg\Bigl(\exists x_{\max}:\ f(x_{\max})=M\Bigr)\iff\forall x:\ f(x)\ne M \]

\[\bigl(f(x)\ne M\bigr)\wedge\bigl(f(x)\le M\bigr)\Rightarrow\bigl(f(x)<M\ \vee\ f(x)>M\bigr)\wedge\neg\bigl(f(x)>M\bigr)\Rightarrow f(x)<M \]

\[\Rightarrow\ \forall x:\ M-f(x)>0\ \Rightarrow\ h:=\frac{1}{M-f(x)}\ \text{良定义}\ \wedge\ C(h)\ \overset{(1)}{\Rightarrow}\ B(h) \]

\[B(h)\Rightarrow\exists k>0\ \forall x:\ \frac{1}{M-f(x)}\le k\Rightarrow f(x)\le M-\frac1k \]

\[\Bigl(\forall x:\ f(x)\le M-\tfrac1k\Bigr)\ \wedge\ \Bigl(\exists x:\ f(x)>M-\tfrac1k\Bigr)\ \Rightarrow\ \bot \]

\[\implies\ \neg\Bigl(\forall x:\ f(x)\ne M\Bigr)\Rightarrow\exists x_{\max}\in[a,b]:\ f(x_{\max})=M \]

最小值(对偶):

\[\Bigl(\forall x:\ f(x)\ne m\Bigr)\wedge\Bigl(\forall x:\ f(x)\ge m\Bigr)\Rightarrow f(x)>m\Rightarrow g:=\frac{1}{f(x)-m},\ C(g)\Rightarrow B(g)\Rightarrow f(x)\ge m+\frac1k \]

\[\Bigl(\forall x:\ f(x)\ge m+\tfrac1k\Bigr)\ \wedge\ \Bigl(\exists x:\ f(x)<m+\tfrac1k\Bigr)\ \Rightarrow\ \bot\ \Rightarrow\ \exists x_{\min}:\ f(x_{\min})=m \]

(或更短:\(C(f)\Rightarrow C(-f)\Rightarrow\max(-f)=-\min f\Rightarrow\exists x_{\min}:\ f(x_{\min})=m\))

  1. 汇成一条析取式(三个反证假设皆推出 \(\bot\))

\[\Bigl(\forall K\ \exists x:\ |f(x)|>K\Bigr)\ \vee\ \Bigl(\forall x:\ f(x)\ne M\Bigr)\ \vee\ \Bigl(\forall x:\ f(x)\ne m\Bigr)\ \Rightarrow\ \bot \]

\[\Rightarrow\ \neg\Bigl(\forall K\ \exists x:\ |f(x)|>K\Bigr)\wedge\neg\Bigl(\forall x:\ f(x)\ne M\Bigr)\wedge\neg\Bigl(\forall x:\ f(x)\ne m\Bigr) \]

\[\iff\ B(f)\ \wedge\ \Bigl(\exists x_{\max}:\ f(x_{\max})=M\Bigr)\ \wedge\ \Bigl(\exists x_{\min}:\ f(x_{\min})=m\Bigr)\ \iff\ B(f)\wedge E(f) \]

定理 B:零点定理 + 介值定理

记(沿用定理 A):\(C(f):\forall t\in[a,b]\ \forall\varepsilon>0\ \exists\delta>0\ \forall x:\bigl(x\in[a,b]\wedge|x-t|<\delta\bigr)\Rightarrow|f(x)-f(t)|<\varepsilon\)

零点定理(Bolzano):

\[C(f)\ \wedge\ \Bigl[\bigl(f(a)<0\ \wedge\ f(b)>0\bigr)\ \vee\ \bigl(f(a)>0\ \wedge\ f(b)<0\bigr)\Bigr]\ \Rightarrow\ \exists c\in(a,b):\ f(c)=0 \]

介值定理:

\[C(f)\ \wedge\ \bigl(f(a)=A\bigr)\ \wedge\ \bigl(f(b)=B\bigr)\ \wedge\ \bigl(A\ne B\bigr)\ \wedge\ \Bigl[(A<D<B)\ \vee\ (B<D<A)\Bigr]\ \Rightarrow\ \exists c\in(a,b):\ f(c)=D \]

归约为零点定理(\(g:=f-D\),只需 \(\vee\) 的一侧;另一侧对 \(A>B\) 同理):

\[C(f)\wedge C(\text{常数 }D)\Rightarrow C(f-D)=C(g) \]

\[\bigl(A<D<B\bigr)\Rightarrow\Bigl(g(a)=A-D<0\Bigr)\ \wedge\ \Bigl(g(b)=B-D>0\Bigr) \]

\[\Bigl(\exists c:\ g(c)=0\Bigr)\ \wedge\ \bigl(g=f-D\bigr)\Rightarrow f(c)=D\ \Rightarrow\ \exists c\in(a,b):\ f(c)=D \]

下面只证零点定理(WLOG 取 \(g(a)<0<g(b)\);或对 \(-g\) 用同一论证):

\[\Bigl(g(a)<0<g(b)\Bigr)\ \vee\ \Bigl(g(b)<0<g(a)\Bigr),\qquad C(g)\ \Rightarrow\ \exists c\in(a,b):\ g(c)=0 \]

局部保号引理(连续性的直接推论,后面两次用到):

\[C(g)\wedge\bigl(g(t)\ne0\bigr)\Rightarrow\exists\eta>0\ \forall x\in[a,b]:\bigl(|x-t|<\eta\bigr)\Rightarrow\Bigl[g(x)\cdot g(t)>0\Bigr] \]

(取 \(\varepsilon=\tfrac{|g(t)|}{2}>0\):\(|g(x)-g(t)|<\tfrac{|g(t)|}{2}\Rightarrow g(x)\) 与 \(g(t)\) 同号)

零点定理的证明(确界原理)

设 \(g(a)<0<g(b)\)。定义

\[E:=\{x\in[a,b]\mid g(x)<0\},\qquad c:=\sup E \]

(1) \(c\) 存在:

\[\bigl(g(a)<0\bigr)\Rightarrow a\in E\Rightarrow E\ne\emptyset;\qquad \bigl(\forall x\in E:\ x\le b\bigr)\Rightarrow b\ \text{为上界} \]

\[E\ne\emptyset\ \wedge\ E\ \text{上有界}\ \overset{\mathbb R\ \text{完备}}{\Rightarrow}\ \exists c=\sup E\in[a,b] \]

(2) \(c>a\)(保号引理于 \(a\)):

\[g(a)<0\Rightarrow\exists\eta>0\ \forall x:\bigl(|x-a|<\eta\wedge x\in[a,b]\bigr)\Rightarrow g(x)<0\Rightarrow x\in E \]

\[\Rightarrow\ \exists x\in E:\ x>a\ \Rightarrow\ c=\sup E\ge x>a \]

(3) \(c<b\)(保号引理于 \(b\)):

\[g(b)>0\Rightarrow\exists\eta>0\ \forall x:\bigl(|x-b|<\eta\bigr)\Rightarrow g(x)>0\Rightarrow \neg\bigl(g(x)<0\bigr)\Rightarrow x\notin E \]

\[\Rightarrow\ (b-\eta,b]\cap E=\emptyset\ \Rightarrow\ \bigl(\forall x\in E:\ x\le b-\eta\bigr)\ \Rightarrow\ c\le b-\eta<b \]

(4) \(g(c)<0\ \vee\ g(c)=0\ \vee\ g(c)>0\)

情形 \(g(c)<0\):

\[g(c)<0\Rightarrow\exists\eta>0\ \forall x:\bigl(|x-c|<\eta\bigr)\Rightarrow g(x)<0 \]

\[\bigl(c<b\bigr)\Rightarrow\exists x:\ c<x<\min\{c+\eta,b\}\Rightarrow\bigl(x\in E\bigr)\ \wedge\ \bigl(x>c=\sup E\bigr)\Rightarrow\bot \]

情形 \(g(c)>0\)(用上确界性质 \(\forall\eta>0\ \exists x\in E:\ x>c-\eta\)):

\[g(c)>0\Rightarrow\exists\eta>0\ \forall x:\bigl(|x-c|<\eta\bigr)\Rightarrow g(x)>0\Rightarrow x\notin E \]

\[\bigl(\exists x\in E:\ c-\eta<x\le c\bigr)\Rightarrow\bigl(|x-c|<\eta\bigr)\Rightarrow x\notin E \]

\[\Rightarrow\ \bigl(x\in E\bigr)\ \wedge\ \neg\bigl(x\in E\bigr)\Rightarrow\bot \]

故

\[\neg\bigl(g(c)<0\bigr)\ \wedge\ \neg\bigl(g(c)>0\bigr)\ \wedge\ \Bigl(g(c)<0\ \vee\ g(c)=0\ \vee\ g(c)>0\Bigr)\ \Rightarrow\ g(c)=0 \]

\[\bigl(g=f-D\bigr)\wedge\bigl(g(c)=0\bigr)\Rightarrow f(c)=D,\qquad a<c<b \]

\[\implies\ \exists c\in(a,b):\ f(c)=D\ ■ \]

推论(与定理 A 合取:连续像必为闭区间)

\[C(f)\Rightarrow\Bigl(\exists x_{\min},x_{\max}:\ m=f(x_{\min})\le f(x)\le f(x_{\max})=M\Bigr)\ \wedge\ \Bigl(\forall D:\ m\le D\le M\Rightarrow\exists c:\ f(c)=D\Bigr) \]

\[\Rightarrow\ f\bigl([a,b]\bigr)=[m,M] \]

(\(m=D\) 或 \(M=D\) 时取 \(c=x_{\min}\) 或 \(x_{\max}\);\(m<D<M\) 时对 \([x_{\min},x_{\max}]\) 或 \([x_{\max},x_{\min}]\) )

定理 C:不动点

记:\(C(f):\forall t\in[a,b]\ \forall\varepsilon>0\ \exists\delta>0\ \forall x:\bigl(x\in[a,b]\wedge|x-t|<\delta\bigr)\Rightarrow|f(x)-f(t)|<\varepsilon\)
自映射条件:\(f:[a,b]\to[a,b]\iff\Bigl(\forall x\in[a,b]:\ a\le f(x)\le b\Bigr)\)

定理:

\[C(f)\ \wedge\ \Bigl(\forall x\in[a,b]:\ a\le f(x)\le b\Bigr)\ \Rightarrow\ \exists c\in[a,b]:\ f(c)=c \]

(注意:结论是 \(c\in[a,b]\),不是 \(c\in(a,b)\)——不动点可落在端点;这是与定理 B 唯一的形似差异。)

证明

令 \(g(x):=f(x)-x\)。

\[C(f)\ \wedge\ C(x\mapsto x)\Rightarrow C(f-x)=C(g) \]

由自映射条件取端点(\(x=a\) 与 \(x=b\) 各一次):

\[\bigl(a\le f(a)\le b\bigr)\Rightarrow g(a)=f(a)-a\ge0,\qquad \bigl(a\le f(b)\le b\bigr)\Rightarrow g(b)=f(b)-b\le0 \]

\[\Rightarrow\ \Bigl(g(a)\ge0\Bigr)\ \wedge\ \Bigl(g(b)\le0\Bigr) \]

三分(对 \(g(a)\)、\(g(b)\) 各用一次,再用 \(\vee\) 并联):

\[\Bigl(g(a)=0\ \vee\ g(a)>0\Bigr)\ \wedge\ \Bigl(g(b)=0\ \vee\ g(b)<0\Bigr) \]

\[\Rightarrow\ \bigl(g(a)=0\bigr)\ \vee\ \bigl(g(b)=0\bigr)\ \vee\ \Bigl(g(a)>0\ \wedge\ g(b)<0\Bigr) \]

情形一(端点即不动点):

\[g(a)=0\Rightarrow f(a)-a=0\Rightarrow f(a)=a\Rightarrow\exists c:=a\in[a,b]:\ f(c)=c \]

\[g(b)=0\Rightarrow f(b)-b=0\Rightarrow f(b)=b\Rightarrow\exists c:=b\in[a,b]:\ f(c)=c \]

情形二(严格异号,套定理 B 零点定理):

\[\Bigl(g(a)>0\ \wedge\ g(b)<0\Bigr)\ \overset{\text{即 }g(a)>0>g(b)}{\wedge}\ C(g)\ \Rightarrow\ \exists c\in(a,b):\ g(c)=0 \]

\[\bigl(g(c)=0\bigr)\Rightarrow f(c)-c=0\Rightarrow f(c)=c \]

\[\bigl(c\in(a,b)\bigr)\Rightarrow\bigl(c\in[a,b]\bigr)\Rightarrow\exists c\in[a,b]:\ f(c)=c \]

合并(\(\vee\) 消去,三种情形结论一致):

\[\Bigl[\exists c=a\Bigr]\ \vee\ \Bigl[\exists c=b\Bigr]\ \vee\ \Bigl[\exists c\in(a,b)\Bigr]\ \Rightarrow\ \exists c\in[a,b]:\ f(c)=c \]

\[\implies\ C(f)\wedge\bigl(f:[a,b]\to[a,b]\bigr)\Rightarrow\exists c\in[a,b]:\ f(c)=c\ ■ \]

退化情形(\(a=b\))

\[a=b\Rightarrow[a,b]=\{a\}\Rightarrow\Bigl(f(a)\in\{a\}\Bigr)\Rightarrow f(a)=a\Rightarrow\exists c:=a:\ f(c)=c \]

(此时不动点存在且唯一,但定理 C 的一般情形不保证唯一性。)

唯一性不成立(只保证 \(\exists\),不保证 \(\exists!\)):

\[\exists c:\ f(c)=c\quad\text{真},\qquad \exists!c:\ f(c)=c\quad\text{未必} \]

\[\neg\Bigl(\forall c_1,c_2:\ \bigl(f(c_1)=c_1\ \wedge\ f(c_2)=c_2\bigr)\Rightarrow c_1=c_2\Bigr) \]

(反例:\(f\equiv x\) 恒等映射,\(\forall x\in[a,b]:\ f(x)=x\),不动点集 \(=[a,b]\),非单点.)

posted @ 2026-11-01 17:53  星灯社小吊扇  阅读(28)  评论(0)    收藏  举报