I'm proving that a $C^r$ function $f:\mathbb{R^2}\rightarrow \mathbb{R}$ is not injective. I've seen various solutions but I want to prove it using techniques related to the Inverse function theorem; because it's a problem from Spivaks "Callculus on Manifolds" text in the chapter on the Inverse function theorem.

I've found a function $g:\mathbb{R^2}\rightarrow \mathbb{R^2}$ defined as $g(x,y)=\bigg(f(x,y),y\bigg)$, and proven that it's $C^r$. To prove my main problem, I'm assuming $f$ is injective, and that allows me to apply the inverse function theorem to $g$, in such a way that gives me an open map from an open set in the domain of $g$ to the open image of $g$, such that $g$ has a differentiable inverse $g^{-1}:g(A)\rightarrow A$.

Now I want to prove a contradiction in the injectivity of $f$, by showing that $g^{-1}$ is not defined at certain points, specifically a line, and this contradicts the differentiability of $g^{-1}$. Consider $f(x,y)=b$ then $g(x,y)=(b,y)$, Now I should be able to find that the line that $g^{-1}$ isn't defined on $\forall (b,z)$, $z\neq y$, since that would imply that $f(x,z)=b$.

My Question I don't quite see how $g^{-1}$ being defined on $(b,z)$ implies that $f(x,z)=b$? $\phantom{}$

Here is the original problem. I've seen other solutions suggest to make $g$ satisfy the conditions of $f$ from $2$-$36$, which is why I left it in. enter image description here



What does it mean for $$ g^{-1}(b, z) = (u, v)? $$ It means that $$ g(u, v) = (b, z). $$ On the other hand, we know that $$ g(u, v) = (f(u, v), v) $$ so setting these two things equal, we see that $$ f(u, v) = b\\ v = z $$ from which we see that $$ f(u, z) = b. $$ I'm not sure if this was what was confusing you, but it seemed to be.

  • $\begingroup$ That's exactly the part that I couldn't get. Thanks! $\endgroup$ – john fowles Jul 1 '18 at 21:19
  • $\begingroup$ Early in my proof I claimed that there must exist a neighbhorood where $D_1f(x,y)\neq 0$ because $D_1f=D_2f=0$ would make the 1-1 condition impossible. Why exactly does $D_1f=D_2f=0$ $\forall$ points in the domain imply that $f$ couldn't be injective? It seems obvious that critical points on a continuous map couldn't be injective in any neighborhood around the point, but how do I express this? $\endgroup$ – john fowles Jul 1 '18 at 21:33
  • $\begingroup$ That sounds like a moment to apply the mean value theorem (or the FTC) along any line in the domain. But I haven't checked this claim, so you're on your own. $\endgroup$ – John Hughes Jul 1 '18 at 22:15
  • $\begingroup$ Ok, Thanks again $\endgroup$ – john fowles Jul 1 '18 at 22:19

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.