隐函数(组)存在定理

(隐函数存在定理-1)
$设二元函数F(x,y)在点(x_0,y_0)的邻域U(x_0,\delta )\times U(y_0,\delta )(\delta >0)内满足以下条件：$
$(1)F(x_0,y_0)=0$；
$(2)F(x,y)、F_y(x,y)在U((x_0,y_0),\delta )(\delta >0)内连续$；
$(3)F_y(x_0,y_0)\ne 0$；
$则\exists {\delta }_0\in (0,\delta )使得在U(x_0,{\delta }_0)内唯一存在满足下述条件的连续函数y=f(x)：$
$(a)y_0=f(x_0)；$
$(b)对\forall x\in U(x_0,{\delta }_0),F(x_0,f(x_0))=0$
$(c)若F_x(x,y)在U((x_0,y_0),\delta )(\delta >0)内连续则y=f(x)存在连续导数，且f^{\prime }(x)=-\frac{F_x}{F_y}。$

证明：

不妨假设$F_y(x_0,y_0)>0$，而$F_y(x_0,y_0)<0$的情形相当于对$F^{\prime }(x,y)=-F(x,y)$进行相关讨论。
又由$F_y(x,y)在U(x_0,\delta )\times U(y_0,\delta )(\delta >0)$内的连续性可知：
$$\exists {\delta }_1,{\delta }_2\in (0,\delta )s.t.对\forall (x,y)\in U(x_0,{\delta }_1)\times U(y_0,{\delta }_2)，均有F_y(x,y)>0$$特别地，对 $\forall y\in U(y_0,{\delta }_2)$均有$F_y(x_0,y)>0$，即：
$$固定x=x_0，则函数F(x_0,y)在y\in U(y_0,{\delta }_2)内单调递增$$又因为，$F(x_0,y_0)=0$，则：$$\{\begin{array}{l}F(x_0,y_0+{\delta }_2)>0\\ F(x_0,y_0-{\delta }_2)<0\end{array}$$进一步由$F(x,y)在U(x_0,\delta )\times U(y_0,\delta )(\delta >0)$内的连续性可知：$$\exists {\delta }_0\in (0,{\delta }_1)s.t.对\forall x\in U(x_0,{\delta }_0)，均有\{\begin{array}{l}F(x,y_0+{\delta }_2)>0\\ F(x,y_0-{\delta }_2)<0\end{array}$$换而言之，$$对任意\tilde{x}\in U(x_0,{\delta }_0)均有F(\tilde{x},y)由负数F(\tilde{x},y-{\delta }_2)连续单调地增长为正数F(\tilde{x},y+{\delta }_2)$$故由零点定理可知：$$对\forall \tilde{x}\in U(x_0,{\delta }_0)均唯一\exists \tilde{y}\in U(y_0,{\delta }_2)s.t.F(\tilde{x},\tilde{y})=0$$上述讨论证明了隐函数$\tilde{y}=f(\tilde{x})$的唯一存在性，接下来说明隐函数的连续性：

取$\hat{x}\in U(x_0,{\delta }_0)$ ，$\hat{y}=f(\hat{x})$，则对充分小 $\epsilon >0$有:$$\{\begin{array}{l}F(\hat{x},\hat{y}+\epsilon )>0\\ F(\hat{x},\hat{y}-\epsilon )<0\end{array}$$又由$F(x,y)$在$(\hat{x},\hat{y}-\epsilon )、(\hat{x},\hat{y}+\epsilon )$处的连续性可知：$$\exists {\delta }^{\prime }\in (0,{\delta }_0)s.t.对\forall x\in U(x_0,{\delta }^{\prime })\subset U(x_0,{\delta }_0),均有\{\begin{array}{l}F(x,\hat{y}+\epsilon )>0\\ F(x,\hat{y}-\epsilon )<0\end{array}⟹\{\begin{array}{l}\hat{y}+\epsilon >f(x)\\ \hat{y}-\epsilon <f(x)\end{array}⟹|\hat{y}-f(x)|<\epsilon$$故,$$\forall \epsilon >0,\exists {\delta }^{\prime }>0,s.t.\forall x\in U(\hat{x},{\delta }^{\prime }):|f(x)-\hat{y}|<\epsilon$$符合隐函数连续性的定义。

下面证明：若$F_x(x,y)$在$U((x_0,y_0),\delta )(\delta >0)$内连续则$y=f(x)$存在连续导数。
任取$\bar{x}\in U(x_0,{\delta }_0)$与充分小的增量$\Delta x，s.t.\bar{x}+\Delta x\in U(x_0,{\delta }_0)$,记：$$\bar{y}=f(\bar{x})，\Delta y=f(\bar{x}+\Delta x)-f(\bar{x})$$由于$F(\bar{x},f(\bar{x}))=0,F(\bar{x}+\Delta x,f(\bar{x}+\Delta x))=0$,则由拉格朗日微分中值定理：$$0=F(\bar{x}+\Delta x,f(\bar{x}+\Delta x))-F(\bar{x},f(\bar{x}))=F(\bar{x}+\Delta x,\bar{y}+\Delta y)-F(\bar{x},\bar{y})=F_x(\bar{x}+\theta \Delta x,\bar{y}+\theta \Delta y)\Delta x+F_y(\bar{x}+\theta \Delta x,\bar{y}+\theta \Delta y)\Delta y(0<\theta <1)$$由于$F_y\ne 0$，则有：$$\frac{\Delta y}{\Delta x}=-\frac{F_x(\bar{x}+\theta \Delta x,\bar{y}+\theta \Delta y)}{F_y(\bar{x}+\theta \Delta x,\bar{y}+\theta \Delta y)}$$由$F_x、F_y与y=f(x)$的连续性：$$\underset{\Delta x⟶0}{\lim }\frac{\Delta y}{\Delta x}=-\frac{F_x(\bar{x},\bar{y})}{F_y(\bar{x},\bar{y})}$$

Remark:
(1) 隐函数存在定理所确定的隐函数只是局部存在；
(2) 隐函数存在定理只给出了隐函数存在的充分条件，若不满足定理条件也可能存在隐函数。

作为上述定理的推广，由类似的推导可以得出：
(隐函数存在定理-2)
$设函数F(\vec{x},y)(\vec{x}\in R^n)在点({\vec{x}}_0,y_0)的邻域U({\vec{x}}_0,\delta )\times U(y_0,\delta )(\delta >0)内满足以下条件：$
$(1)F({\vec{x}}_0,y_0)=0$；
$(2)F(\vec{x},y)、F_y(\vec{x},y)在U({\vec{x}}_0,\delta )\times U(y_0,\delta )(\delta >0)内连续$；
$(3)F_y({\vec{x}}_0,y_0)\ne 0$；
$则\exists {\delta }_0\in (0,\delta )使得在U({\vec{x}}_0,{\delta }_0)内唯一存在满足下述条件的连续函数y=f(\vec{x})：$
$(a)y_0=f({\vec{x}}_0)；$
$(b)对\forall \vec{x}\in U({\vec{x}}_0,{\delta }_0),F({\vec{x}}_0,f({\vec{x}}_0))=0$
$(c)若F(\vec{x},y)在U({\vec{x}}_0,\delta )\times U(y_0,\delta )(\delta >0)内存在关于各x_i的连续偏导数，那么y=f(\vec{x})存在关于各x_i的连续偏导数，且\frac{\partial f(\vec{x})}{\partial x_i}=-\frac{F_{x_i}}{F_y}(i=1,2,\dots ,n)。$

(隐函数组存在定理)
$设向量函数\vec{F}(\vec{x},\vec{u})=(F_1(\vec{x},\vec{u}),F_2(\vec{x},\vec{u}),\dots ,F_m(\vec{x},\vec{u})),其中\vec{x}=(x_1,x_2,\dots ,x_n)\in R^n,\vec{u}=(u_1,u_2,\dots ,u_m)\in R^m,在U({\vec{x}}_0,\delta )\times U({\vec{u}}_0,\delta )(\delta >0)内满足条件：$
$(1)F_i({\vec{x}}_0,{\vec{u}}_0)=0，i=1,2,\dots ,m；$
$(2)F_i(\vec{x},\vec{u})与其关于u_i各个偏导在U({\vec{x}}_0,\delta )\times U({\vec{u}}_0,\delta )内连续；$
$(3)雅可比行列式\frac{\partial (F_1,F_2,\dots ,F_m)}{\partial (u_1,u_2,\dots ,u_m)}{|}_{\vec{x}={\vec{x}}_0;\vec{u}={\vec{u}}_0}\ne 0$
$则\exists {\delta }_0\in (0,\delta )使得在U({\vec{x}}_0,{\delta }_0)\times U({\vec{u}}_0,{\delta }_0)内唯一存在m维n元向量函数$$$\vec{u}(\vec{x})=(u_1(\vec{x}),u_2(\vec{x}),\dots ,u_m(\vec{x}))$$满足：
$(a){\vec{u}}_0=(u_1({\vec{x}}_0),u_2({\vec{x}}_0),\dots ,u_m({\vec{x}}_0))；$
$(b)对任意F_i与\vec{x}\in U({\vec{x}}_0,{\delta }_0)均有F_i(\vec{x},\vec{u}(\vec{x}))=0；$
$(c)若F_i对任意x_j有连续的偏导数，则u_k对x_j的偏导连续，记$
$$A={\left[\begin{array}{cccc}\frac{\partial F_1(\vec{x},\vec{u})}{\partial x_1}& \frac{\partial F_1(\vec{x},\vec{u})}{\partial x_2}& \dots & \frac{\partial F_1(\vec{x},\vec{u})}{\partial x_n}\\ \frac{\partial F_2(\vec{x},\vec{u})}{\partial x_1}& \frac{\partial F_2(\vec{x},\vec{u})}{\partial x_2}& \dots & \frac{\partial F_2(\vec{x},\vec{u})}{\partial x_n}\\ ⋮& ⋮& & ⋮\\ \frac{\partial F_m(\vec{x},\vec{u})}{\partial x_1}& \frac{\partial F_m(\vec{x},\vec{u})}{\partial x_2}& \dots & \frac{\partial F_m(\vec{x},\vec{u})}{\partial x_n}\end{array}\right]}_{m\times n}B={\left[\begin{array}{cccc}\frac{\partial F_1(\vec{x},\vec{u})}{\partial u_1}& \frac{\partial F_1(\vec{x},\vec{u})}{\partial u_2}& \dots & \frac{\partial F_1(\vec{x},\vec{u})}{\partial u_m}\\ \frac{\partial F_2(\vec{x},\vec{u})}{\partial u_1}& \frac{\partial F_2(\vec{x},\vec{u})}{\partial u_2}& \dots & \frac{\partial F_2(\vec{x},\vec{u})}{\partial u_m}\\ ⋮& ⋮& & ⋮\\ \frac{\partial F_m(\vec{x},\vec{u})}{\partial u_1}& \frac{\partial F_m(\vec{x},\vec{u})}{\partial u_2}& \dots & \frac{\partial F_m(\vec{x},\vec{u})}{\partial u_m}\end{array}\right]}_{m\times m}$$则$${\vec{u}}^{\prime }(\vec{x})={\left[\begin{array}{cccc}\frac{\partial u_1(\vec{x})}{\partial x_1}& \frac{\partial u_1(\vec{x})}{\partial x_2}& \dots & \frac{\partial u_1(\vec{x})}{\partial x_n}\\ \frac{\partial u_2(\vec{x})}{\partial x_1}& \frac{\partial u_2(\vec{x})}{\partial x_2}& \dots & \frac{\partial u_2(\vec{x})}{\partial x_n}\\ ⋮& ⋮& & ⋮\\ \frac{\partial u_m(\vec{x})}{\partial x_1}& \frac{\partial u_m(\vec{x})}{\partial x_2}& \dots & \frac{\partial u_m(\vec{x})}{\partial x_n}\end{array}\right]}_{m\times n}=-B^{-1}A$$

证明：（数学归纳法）
当$m=2$时，由于$$\frac{\partial (F_1,F_2)}{\partial (u_1,u_2)}{|}_{\vec{x}={\vec{x}}_0;\vec{u}={\vec{u}}_0}={\left|\begin{array}{cc}\frac{\partial F_1}{\partial u_1}& \frac{\partial F_1}{\partial u_2}\\ & \\ \frac{\partial F_2}{\partial u_1}& \frac{\partial F_2}{\partial u_2}\end{array}\right|}_{\vec{x}={\vec{x}}_0;\vec{u}={\vec{u}}_0}\ne 0$$则其中各行不可能所有元素为零，不妨假设$\frac{\partial F_2}{\partial u_2}\ne 0$。若$\frac{\partial F_2}{\partial u_2}=0$，则$\frac{\partial F_2}{\partial u_1}$必不为零，同样可进行如下相同的处理：
由隐函数存在定理-2：$\exists {\delta }_1\in (0,\delta )$使得方程$F_2(\vec{x},u_1,u_2)=0$在$U({\vec{x}}_0,{\delta }_1)\times U({u_1}^0,{\delta }_1)$内存在唯一的连续函数（其关于$u_1$偏导也连续）$$u_2=u_2(\vec{x},u_1),且\frac{\partial u_2}{\partial u_1}=-\frac{\frac{\partial F_2}{\partial u_1}}{\frac{\partial F_2}{\partial u_2}}$$代入另一方程得：
$$H(\vec{x},u_1)=F_1(\vec{x},u_1,u_2(\vec{x},u_1))=0$$由于
$$\frac{\partial H({\vec{x}}_0,{u_1}^0)}{\partial u_1}=\frac{\partial F_1}{\partial u_1}+\frac{\partial F_1}{\partial u_2}\frac{\partial u_2}{\partial u_1}=\frac{\partial F_1}{\partial u_1}-\frac{\partial F_1}{\partial u_2}\frac{\frac{\partial F_2}{\partial u_1}}{\frac{\partial F_2}{\partial u_2}}=\frac{1}{\frac{\partial F_2}{\partial u_2}}(\frac{\partial F_2}{\partial u_2}\frac{\partial F_1}{\partial u_1}-\frac{\partial F_1}{\partial u_2}\frac{\partial F_2}{\partial u_1})=\frac{1}{\frac{\partial F_2}{\partial u_2}}\frac{\partial (F_1,F_2)}{\partial (u_1,u_2)}\ne 0$$则$\exists {\delta }_0\in (0,{\delta }_1)$使得方程$H(\vec{x},u_1)=0$在$U({\vec{x}}_0,{\delta }_0)$内存在唯一连续函数$u_1=u_1(\vec{x})$故该方程组
$$\{\begin{array}{l}{F}_1(\vec{x},u_1,u_2)=0\\ F_2(\vec{x},u_1,u_2)=0\end{array}确定了连续的隐函数组\{\begin{array}{l}{u}_1=u_1(\vec{x})\\ u_2=u_2(\vec{x},u_1(\vec{x}))\end{array}$$
进一步，若$F_i(\vec{x},\vec{u})（i=1,2,...,m）$对$x_j（j=1,2,\dots ,n）$均有连续偏导数
$$\{\begin{array}{l}\frac{\partial F_1}{\partial x_i}+\frac{\partial F_1}{\partial u_1}\frac{\partial u_1}{\partial x_i}+\frac{\partial F_1}{\partial u_2}\frac{\partial u_2}{\partial x_i}=0\\ \\ \frac{\partial F_2}{\partial x_i}+\frac{\partial F_2}{\partial u_1}\frac{\partial u_1}{\partial x_i}+\frac{\partial F_2}{\partial u_2}\frac{\partial u_2}{\partial x_i}=0\end{array}⟺\left[\begin{array}{cc}\frac{\partial F_1}{\partial u_1}& \frac{\partial F_1}{\partial u_2}\\ & \\ \frac{\partial F_2}{\partial u_1}& \frac{\partial F_2}{\partial u_2}\end{array}\right]\left[\begin{array}{c}\frac{\partial u_1}{\partial x_i}\\ \\ \frac{\partial u_2}{\partial x_i}\end{array}\right]=-\left[\begin{array}{c}\frac{\partial F_1}{\partial x_i}\\ \\ \frac{\partial F_2}{\partial x_i}\end{array}\right]⟺\left[\begin{array}{cc}\frac{\partial F_1}{\partial u_1}& \frac{\partial F_1}{\partial u_2}\\ & \\ \frac{\partial F_2}{\partial u_1}& \frac{\partial F_2}{\partial u_2}\end{array}\right]\left[\begin{array}{cccc}\frac{\partial u_1}{\partial x_1}& \frac{\partial u_1}{\partial x_2}& \dots & \frac{\partial u_1}{\partial x_n}\\ & & & \\ \frac{\partial u_2}{\partial x_1}& \frac{\partial u_2}{\partial x_2}& \dots & \frac{\partial u_2}{\partial x_n}\end{array}\right]=-\left[\begin{array}{cccc}\frac{\partial F_1}{\partial x_1}& \frac{\partial F_1}{\partial x_2}& \dots & \frac{\partial F_1}{\partial x_n}\\ & & & \\ \frac{\partial F_2}{\partial x_1}& \frac{\partial F_2}{\partial x_2}& \dots & \frac{\partial F_2}{\partial x_n}\end{array}\right]⟺{\vec{u}}^{\prime }(\vec{x})=-B^{-1}A$$
假设对$m-1$个方程构成的方程组定理均成立，则$m$个方程构成的方程组经过一次消元后得到$m-1$个方程构成的方程组，可推知定理成立，此处不再赘述。