Implicit function Theorem: In the general implicit function theorem for $m$ variables and $m$ implicit equations in the form $$\begin{align} \mathbf F(x_1,x_2,\ldots,x_n, u_1, u_2, \ldots, u_m) = 0 \end{align}$$ where $\mathbf F=\langle F_1, F_2,...,F_m \rangle$

I have been introduced to the requirement that the square jacobian matrix for $\mathbf F(u_1,...,u_n)$ must be invertible which means the determinant should be non zero. This is apparently analogous to requiring $\frac{\partial f}{\partial y}\ne0$ for the $2D$ case $F(x,y)=0$.

Can someone please explain any intuition behind this requirement?


Consider the warm-up exercise, of formulating the implicit function theorem for linear functions. That is, where $\mathbf F(x,u)=0$ can be given by a matrix multiplication formula like $$ Ax+Cu=b$$ where $x$ in an $n$-vector and $u$ an $m$-vector, with matrices $A$ and $C$ and fixed vector $b$. Linear algebra tells us this equation is always uniquely solveable for $u$ given $x$ precisely when the $m\times m$ matrix $C$ is non-singular, that is, has non-vanishing determinant.

In this case $C$ is the Jacobian.

Now in the non-linear but differentiable case. A differentiable function is, intuitively, one which is well approximated by a linear one. One might expect that what holds in the exactly linear case carries over to the differentiable case. The technical content of the implicit function theorem is that this is so.

| cite | improve this answer | |
  • $\begingroup$ Thank you for this! It makes sense. $\endgroup$ – user523384 Mar 23 '19 at 9:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.