Exercises
[1GZ] In the same assumptions as the previous theorem [1GD], show that there exist \(\varepsilon {\gt}0\) and a continuous function \(\tilde g:Vβ β\) where \(I=(\overline a-\varepsilon ,\overline a+\varepsilon )\) and \(V=U'Γ I\) is open in \(β^ n\), such that
\begin{equation} β (x',a)β V \quad ,\quad (x',\tilde g(x',a))β U \quad \text{e} \quad f\big(x',\tilde g(x',a)\big)=a \quad .\label{eq:tilde_ g_ teor_ funz_ inv} \end{equation}54Vice versa if \(xβ U\) and \(a=f(x)\) and \(aβ I\) then \(x_ n=\tilde g(x',a)\).
Note that the previous relation means that, for each fixed \(x'β U'\), the function \(\tilde g(x',β )\) is the inverse of the function \(f(x',β )\) (when defined on appropriate open intervals).
So, moreover, the function \(\tilde g\) is always differentiable with respect to \(a\), and the partial derivative is
\[ \frac{\partial ~ }{\partial {a}} \tilde g(x',a)=\frac{1}{\frac{\partial ~ }{\partial {x_ n}} f(x',\tilde g(x',a))}\quad . \]The other derivatives instead (obviously) are as in the theorem [1GD].
The regularity of \(\tilde g\) is the same as \(g\): if \(f\) is Lipschitz then \(\tilde g\) is Lipschitz; if \(fβ C^ k(U)\) then \(\tilde gβ C^ k(V)\).
1