🧮 Brain Teaser

The Nowhere-Zero Derivative Paradox

Let f:RRf: \mathbb{R} \to \mathbb{R} be differentiable everywhere, and suppose f(x)0f'(x) \neq 0 for all xRx \in \mathbb{R}.

Must ff' have constant sign? That is, must either f(x)>0f'(x) > 0 for all xx, or f(x)<0f'(x) < 0 for all xx?

Prove or find a counterexample.

Darboux's theoremintermediate value propertyderivativesreal analysis

Answer: Nowhere-Zero Derivative Paradox

Key Idea / Intuition

Your first instinct might be: "If ff' is never zero, it can't switch sign, because by the Intermediate Value Theorem it would have to pass through zero." And this is exactly right — but the key insight is subtle: the derivative ff' need not be continuous, yet it still satisfies the Intermediate Value Property (Darboux's theorem). So even though ff' might be discontinuous, it cannot skip over zero. Therefore yes: ff' must have constant sign.


Formal Proof / Solution

Claim: If f:RRf: \mathbb{R} \to \mathbb{R} is differentiable everywhere and f(x)0f'(x) \neq 0 for all xx, then ff' has constant sign.

Step 1: State Darboux's Theorem.

Darboux's Theorem: If ff is differentiable on [a,b][a, b], then ff' satisfies the Intermediate Value Property: for any a<ba < b and any value kk strictly between f(a)f'(a) and f(b)f'(b), there exists c(a,b)c \in (a, b) with f(c)=kf'(c) = k.

This is remarkable because ff' need not be continuous — yet it cannot "jump" over any value. (The proof uses that g(x)=f(x)kxg(x) = f(x) - kx achieves its extremum at an interior point where g=0g' = 0.)

Step 2: Apply Darboux to conclude constant sign.

Suppose for contradiction that ff' takes both positive and negative values. Then there exist a,bRa, b \in \mathbb{R} with f(a)>0f'(a) > 0 and f(b)<0f'(b) < 0.

By Darboux's theorem applied on the interval [min(a,b),max(a,b)][\min(a,b),\, \max(a,b)], since 00 lies strictly between f(a)f'(a) and f(b)f'(b), there exists cc between aa and bb such that: f(c)=0.f'(c) = 0.

This contradicts the assumption that f(x)0f'(x) \neq 0 for all xx.

Conclusion: ff' cannot change sign. Since ff' is never zero, we must have either f(x)>0f'(x) > 0 for all xRx \in \mathbb{R}, or f(x)<0f'(x) < 0 for all xRx \in \mathbb{R}. \blacksquare


Why this is surprising:

One might think: "Maybe ff' oscillates wildly enough to avoid zero while still changing sign — like sin(1/x)\sin(1/x)." But Darboux blocks this entirely. The IVP for derivatives is a hidden rigidity that survives even in the absence of continuity.

A classic example of a discontinuous derivative: f(x)=x2sin(1/x)f(x) = x^2 \sin(1/x) (with f(0)=0f(0) = 0) has f(0)=0f'(0) = 0 but ff' oscillates near 00. This shows derivatives can be wild — but Darboux guarantees they can never skip a value.

Type: analysisEdit on GitHub ↗