6
$\begingroup$

Let $f$ be a continuous function on the interval $[a,b]$ such as for all x in $(a,b)$ :
The right hand derivative equals $0$.
Is $f$ constant?

The obvious reflex one would have is to try to find a counter-example, I tried, many times and I always failed.
I assumed it to be true and tried to prove it, I tried manipulating sequences, it never worked.
I tried calculating the left hand derivative of a point by using a sequence, it seemed to work at first, but it didn't as well.
All in all, I think $f$ may not be necessarily constant.

$\endgroup$
5
  • 2
    $\begingroup$ Doesn't differentiable imply right hand derivative = left hand derivative? $\endgroup$ Jan 13, 2020 at 22:00
  • $\begingroup$ Oops I made a mistake, the function isn't differentiable. $\endgroup$
    – Lazarus
    Jan 13, 2020 at 22:01
  • $\begingroup$ Try $f(x) = 1$ when $x\geq0$ and $f(x) = 0$ otherwise. $\endgroup$
    – irchans
    Jan 13, 2020 at 22:09
  • 1
    $\begingroup$ on which interval? and would it even be continuous? $\endgroup$
    – Lazarus
    Jan 13, 2020 at 22:11
  • $\begingroup$ Oops, my $f$ was not continuous! $\endgroup$
    – irchans
    Jan 14, 2020 at 21:11

2 Answers 2

5
$\begingroup$

Yes, $f$ is necessarily constant.

One can proceed similarly as in the proofs of the mean value theorem and Rolle's theorem for differentiable functions.

It suffices to show that $f(c) = f(d)$ for $a < c < d < b$. The continuity of $f$ then implies that $f$ is constant on $[a, b]$.

Assume that $f(c) \ne f(d)$, without loss of generality $f(c) < f(d)$. Consider the function $$ g(x) = f(x) - (x-c)\frac{f(d)-f(c)}{d-c} $$ for $x \in [c, d]$. The right-hand derivative of $g$ is $$ g_+'(x) = f_+'(x) - \frac{f(d)-f(c)}{d-c} = - \frac{f(d)-f(c)}{d-c} < 0 $$ for all $x \in [c, d)$.

But $g(c) = g(d)$, so that $g$ attains its minimum at a point $x_0 \in [c, d)$, where $$ g_+'(x_0) = \lim_{x \to x_0^+} \frac{g(x)-g(x_0)}{x-x_0} \ge 0 \, , $$ which is a contradiction. This completes the proof.

$\endgroup$
1
$\begingroup$

Given $\epsilon>0$, we find for each $x\in[a,b)$, the set $$ A_x:=\{\,\xi\in[a,b]\mid \xi>x,\left|\tfrac{f(\xi)-f(x)}{\xi-x}\right|\le \epsilon \,\}$$ contains some open interval $(x,y)$. Let $y^*(x)$ be the supremum of all $y$ with $(x,y)\subseteq A_x$. Clearly, $y^*(x)>x$. By continuity, $y^*(x)\in A_x$.

Suppose $y^*(a)<b$. Then for $y^*(a)<\xi<y^*(y^*(a))$, we have $$\frac{f(\xi)-f(a)}{\xi-a}=\frac{\xi-y^*(a)}{\xi-a}\cdot\frac{f(\xi)-f(y^*(a))}{\xi-y^*(a)}+ \frac{y^*(a)-a}{\xi-a}\cdot\frac{f(y^*(a))-f(a)}{y^*(a)-a},$$ which is a convex combination of numbers $\in[-\epsilon,\epsilon]$. We conclude that $\xi\in A_a$. It follows that $y^*(a)\ge y^*(y^*(a))$, contradiction. We conclude that $y^*(a)=b$, so $A_a=(a,b]$. As $\epsilon$ was arbitrary, $f$ must be constant.

$\endgroup$
1
  • $\begingroup$ could you clarify what do you mean by a convex combination of numbers? $\endgroup$
    – Lazarus
    Jan 13, 2020 at 22:40

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.