Theorem: Let $I\subset \mathbb R$ be an interval and a function $f:I \to \mathbb R$ be strictly monotone and continuous on $I$.Let $g:=f^{-1}:f(I)\to \mathbb R$.If $f$ is differentiable at $c\in I$ and $f'(c)\neq0$ then $g$ is differentiable at $f(c)$ and $g'(f(c))=\frac{1}{f'(c)}$ .

In this theorem can we replace 'strictly monotone and continuous' by the assumptions that $f$ is 'invertible and continuous' or by 'injective and continuous' or by 'injective and having IVP on $I$'.

I think any of these conditions will act the same i.e. they will all provide me with the facts I need to prove the original version of the theorem. Another question is how to generalize this theorem for arbitrary domain $X$.


As an example I'll consider an invertible and continuous function and will show that it has to be strictly monotonic.

Assume $x<y$ are points in $I$ and, w.l.o.g, let's assume $f(x) < f(y)$. ($f(x)=f(y)$ would contradict the fact that $f$ is invertible).

Now assume $x < r < y$. Claim: $f(x) < f(r) < f(y)$. Assume that this is not true. The $f(x)\ge f(r) $ of $f(r)\ge f(y)$. Since both cases are similar, I'll only look at the first one. $f(x)=f(r)$ is impossible, so $f(x) > f(r) $. But then be the mean value theorem there is $s \in (r,y)$ sucht that $f(s) = f(x)$, contradicting the invertibility of $f$.

  • $\begingroup$ So,as I am saying any of these condition is fine,right? $\endgroup$ – Kishalay Sarkar Dec 23 '19 at 9:05
  • $\begingroup$ @KishalaySarkar yep. But you need to work on connected sets, i.e. intervals. $\endgroup$ – Thomas Dec 23 '19 at 9:05
  • $\begingroup$ And the fact $f'(c)\neq 0$ is essential because we want to avoid the situation of infinite derivative of inverse(although it is a bit informal to say,because we do not consider the differential quotient's infinite limit as a derivative).e.g.$ f(x)=x^3$ with $f^{-1}(x)=x^{\frac{1}{3}}$. $\endgroup$ – Kishalay Sarkar Dec 23 '19 at 9:07
  • $\begingroup$ @KishalaySarkar Right. That condition is, of course, still needed. $\endgroup$ – Thomas Dec 23 '19 at 9:09
  • $\begingroup$ How to generalize this theorem for an arbitrary domain $X$ in place of interval domain $I$. $\endgroup$ – Kishalay Sarkar Dec 23 '19 at 9:48

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.