Ngô Quốc Anh

June 3, 2022

Non-zero derivative does not imply the continuity of the inverse

Filed under: Uncategorized — Ngô Quốc Anh @ 0:20

A classical version of the inverse function theorem for 1D asserts the following.

Suppose that f : U \to V is an invertible function with inverse f^{-1} : V \to U. If f is differentiable at x_0 with f'(x_0) \ne 0 and f^{-1} is continuous at y_0 = f(x_0), then f^{-1} is differentiable at y_0 and

\displaystyle \big(f^{-1}\big)'(y_0) = \frac 1{f'(x_0)}.

Very often, some of the above hypotheses for f is replaced by, say, f is continuous “in a neighborhood” and one-to-one. But in the proof, while the one-to-one property immediately gives the existence of the inverse function f^{-1}, the “non-local” continuity yields the continuity of the inverse f^{-1}. Here we are interested in the “local” continuity/differentiablity.

In this note, we show by constructing counter-examples that in general the continuity of f^{-1} at y_0 cannot be dropped. In the sequel, we denote by \mathbb N the set set of positive integers, namely \mathbb N = \{1,2,3,...\}.

1. The case of continuity

We start with the simplest case, namely if we only assume that f is continuous at x_0. In this scenario, the following example due to R.E. Megginson is well-known. Consider the function f : \mathbf R \to \mathbf R defined by

\displaystyle f(x)=\left\{\begin{aligned} & \frac 1{2n} & & \text{if  } x= \frac 1n \text{ for some } n \in \mathbb N \\  & \frac 1n & & \text{if  } x= n \text{ for some odd } n \in \mathbb N \setminus \{1\} \\ & \frac n2 & & \text{if  } x= n \text{ for some even } n \in \mathbb N \setminus \{1\} \\ & x & & \text{otherwise}.\end{aligned}\right.

It is not hard to see that f is well-defined and invertiable. Moreover, f is continous at x=0. However, as f(0)=0, its inverse function f^{-1} fails to be continous at 0 because

\displaystyle f^{-1} \Big( \frac 1{2n+1}\Big) = 2n+1 \nearrow +\infty

as n \to +\infty. This concludes the construction.

2. The case of zero derivative

We now modify the above example to handle the case of differentiable functions at the given point. A careful calculation shows that in the above example, the funtion f is not differentiable at x=0 because

\displaystyle \lim_{n \to +\infty} \frac{f (1/n) - f(0)}{1/n} = \frac 12 \ne 1 = \lim_{n \to +\infty}\frac{f (-1/n) - f(0)}{-1/n} .

Toward a new construction, let us denote

\phi : x \mapsto x^3,

which is one-to-one. Having the function \phi, we construct the desired function f as follows.

\displaystyle f(x)=\left\{\begin{aligned} & \phi (x/2) & & \text{if  } x= 1/n \text{ for some } n \in \mathbb N \\  & \phi (1/x) & & \text{if  } x= n \text{ for some odd } n \in \mathbb N \setminus \{1\} \\ & \phi (x/2) & & \text{if  } x= n \text{ for some even } n \in \mathbb N \setminus \{1\} \\ & \phi (x) & & \text{otherwise}.\end{aligned}\right.

It is not hard to see that f is well-defined and invertiable. We now show that f is differentiable at 0 with f'(0)=0. To see this, we just consider f(x) for |x|<1. From the definition of f and the monotonicity of \phi we note that

\displaystyle |f(x)| \leq |\phi(x)| \leq x^2

for any |x|<1. Hence

\displaystyle 0 \leq \lim_{x \to 0} \Big| \frac{f(x) - f(0)}{x}\Big| = \lim_{x \to 0} \Big| \frac{f(x) }{x}\Big| \leq \lim_{x \to 0} |x| = 0.

This shows that f is differentiable at x=0 with f'(0)=0. Arguing as in the case of continuity, we know that

\displaystyle f^{-1} \Big( \phi \big(\frac 1{2n+1} \big)\Big) = 2n+1 \nearrow +\infty

and

\displaystyle  \phi \big(\frac 1{2n+1} \big) = \frac 1{(2n+1)^3} \searrow 0

as n \to +\infty. This shows that f^{-1} is not continuous at 0.

3. The case of non-zero derivative

We are now arrive in the most delicated case. For simplicity, let us construct some bijective mapping f : \mathbf R \to f(\mathbf R). The following example is due to Mr. Nghia. The desired function f is given as follows

\displaystyle f(x)=\left\{\begin{aligned} & \frac 1{(n+1)^2 + k} & & \text{if  } x= \frac 1{n^2+k} \text{ for some } n \in \mathbb N \text{ and some } k \in \{0,...,2n\} \\ & \frac 1x & & \text{if  } x \in \{ (n+1)^2+2n+1, (n+1)^2 + 2n+2 \} \text{ for some } n \in \mathbb N \\ & x & & \text{otherwise}.\end{aligned}\right.

It is now standard to verify that f is well-defined and invertiable. Obviously, f(\mathbf R) is not the whole \mathbf R because 1/3 \notin f(\mathbf R). However, there holds (-1/4, 1/4) \subset f(\mathbf R) and f(0)=0. We now examine f at 0 and its inverse f^{-1} at 0. We now show that f is differentiable at 0 with f'(0)=1. Again, we just consider f(x) for |x|<1. Clearly, it suffices to show that

\displaystyle \lim_{n \to +\infty} \frac{f (x_n) - f(0)}{x_n} = 1

for any sequence (x_n) converging to 0. This is done once we can show that both

\displaystyle \lim_{n \to +\infty} \frac{f (y_n) - f(0)}{y_n} = 1 = \lim_{n \to +\infty} \frac{f (z_n) - f(0)}{z_n}

hold true, where z_n=1/n and (y_n) is any sequence having no term 1/n. This is because given arbitrary \varepsilon>0, there are two positive integers N_1 and N_2 such that

\displaystyle \Big|\frac{f (y_n) - f(0)}{y_n} -1 \Big| < \varepsilon \quad \forall n \geq N_1

and

\displaystyle \Big|\frac{f (z_n) - f(0)}{z_n} -1 \Big| < \varepsilon \quad \forall n \geq N_2.

Simply taking N = \max\{N_1, N_2\} we conclude that the above estimate holds for all n \geq N regardless of either x_n \ne 1/n or x_n=1/n. Going back to the limits while the case of y_n is trivial by the definition of f, the case z_n = 1/n is also simple because in this scenario, we easily get

\displaystyle \frac 1n = \frac 1{m^2+k}

for some m \geq 1 and some 0 \leq k \leq 2m. Hence

\displaystyle \frac{m^2}{(m+2)^2} \leq \frac{f (z_n) - f(0)}{z_n} = \frac{m^2+k}{(m+1)^2+k} \leq \frac{m^2+2m}{(m+1)^2}.

As n \to +\infty, we must have m \to +\infty. Hence from the above two-sided estimate we must have

\displaystyle \lim_{n \to +\infty} \frac{f (z_n) - f(0)}{z_n} = 1.

Hence

\displaystyle \lim_{n \to +\infty} \frac{f (x_n) - f(0)}{x_n} = 1

as claimed. Thus, we have just shown that f is differentiable at 0 with f'(0)=1. Finally, we show that the inverse function f^{-1} is not continuous at 0. Indeed, this follows from

\displaystyle f^{-1} \Big( \frac 1{(n+1)^2 + 2n+2}\Big) = (n+1)^2 + 2n+2 \nearrow +\infty

as n \to +\infty.

Our last observation concerns the case when f is twice differentiable at x_0. However, in this case f is differentiable in a neighborhood of x_0, which is enough to conclude that its inverse f^{-1} must be continuous at y_0.

Leave a Comment »

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Create a free website or blog at WordPress.com.