Rolle's Theorem

Throw a ball straight up into the air. It rises, slows down, stops for an instant, and falls back. At the very top — the moment it is neither rising nor falling — its velocity is exactly zero. Not approximately zero. Exactly zero.

You did not need calculus to know that. You knew it from experience, from watching every ball you have ever thrown. But here is the mathematical version of the same observation. The ball's height is a smooth function of time. At the moment of the throw it is at some height, and some time later — when it returns to the same height — it is back to the same value. Somewhere between those two moments, the height function stopped increasing and started decreasing. At that transition point, the rate of change of height was zero. The derivative was zero.

This is the simplest and one of the most powerful ideas in calculus. It is called Rolle's theorem, and it is the foundation on which the entire theory of the Mean Value Theorem is built.

The idea, drawn

Take any smooth curve that starts and ends at the same height. It doesn't matter what happens in between — the curve can go up, come down, wiggle, loop around — as long as it starts at height f(a) and ends at the same height f(b) = f(a), and the curve has no sharp corners and no breaks in between.

Look at such a curve. Somewhere between a and b, the curve must have a highest point or a lowest point (or both). At that extreme point, the tangent line to the curve is horizontal — its slope is zero.

That is the entire content of Rolle's theorem. If the curve connects two points at the same height, without breaks or corners, then somewhere in between the derivative must be zero.

Why must there be a highest or lowest point? Because a continuous function on a closed interval always achieves its maximum and minimum values — that is the Extreme Value Theorem. The curve cannot just drift upward forever; it is trapped between a and b, and continuity forces it to turn around somewhere. At that turning point, the tangent is flat.

The formal statement

Reading the conditions. Each condition is doing real work:

• Continuous on [a, b] ensures there are no jumps or gaps. Without this, the curve could leap from one height to another without ever passing through the intermediate values, and there might be no turning point at all.

• Differentiable on (a, b) ensures there are no sharp corners. The function |x| on [-1, 1] is continuous, satisfies f(-1) = f(1) = 1, and does have a minimum at x = 0 — but the tangent at x = 0 is not defined (it is a corner, not a smooth peak). Rolle's theorem does not apply to such functions.

• f(a) = f(b) is the trigger. Without equal endpoint values, the theorem makes no guarantee. (The generalisation to unequal endpoints is Lagrange's Mean Value Theorem, which you will meet in the next article.)

Notice that the theorem says "there exists at least one." There could be many such points. If the curve goes up and down multiple times between a and b, it may have several horizontal tangents. The theorem only promises at least one.

The proof

The proof uses one fact from analysis: a continuous function on a closed interval attains its maximum and minimum values. (This is the Extreme Value Theorem.) Given that, the argument is clean.

Case 1: f is constant on [a, b]. If f(x) = f(a) for every x in [a, b], then f'(x) = 0 for every x in (a, b). Any point in (a, b) works as c. Done.

Case 2: f is not constant. Since f is not constant and f(a) = f(b), the function must take some value different from f(a) at some point in the interior. Two sub-cases arise:

• If f takes a value greater than f(a) somewhere in (a, b), then by the Extreme Value Theorem, f attains its maximum value at some point c \in [a, b]. Since the maximum is strictly greater than f(a) = f(b), the point c cannot be an endpoint — it must lie in the open interval (a, b). Now, f has a local maximum at c, and f is differentiable at c. At a local maximum of a differentiable function, the derivative must be zero. (If f'(c) > 0, the function would be increasing at c and could not be at a maximum; if f'(c) < 0, the function would be decreasing and again could not be at a maximum.) So f'(c) = 0.

• If f takes a value less than f(a) somewhere in (a, b), the same argument works with the minimum instead of the maximum.

In either case, a point c in (a, b) with f'(c) = 0 exists. \blacksquare

The key step — "at a local extremum of a differentiable function, f'(c) = 0" — is sometimes called Fermat's lemma (it is a small observation, not a deep theorem). If you have read the article on critical points, you have already seen this reasoning in action.

Why each condition matters

Each of the three conditions is necessary. Remove any one, and the conclusion can fail. Here are the standard counterexamples.

Without continuity. Define f on [0, 1] as: f(x) = x for 0 \le x < 1, and f(1) = 0. Then f(0) = f(1) = 0, and f is differentiable on (0, 1) with f'(x) = 1 everywhere. The derivative is never zero. The problem is the jump at x = 1: f is not continuous on the closed interval.

Without differentiability. Take f(x) = 1 - |x| on [-1, 1]. Then f(-1) = f(1) = 0, and f is continuous on [-1, 1]. The function reaches its maximum at x = 0, but f'(0) does not exist — there is a sharp corner. The tangent is not defined, so the conclusion "f'(c) = 0" cannot even be stated.

Without f(a) = f(b). Take f(x) = x on [0, 1]. Then f(0) = 0 and f(1) = 1 — different values. The function is continuous and differentiable everywhere, but f'(x) = 1 for all x. The derivative is never zero. The equal-endpoints condition is essential.

Applications

Rolle's theorem is not just a theoretical curiosity. It is a sharp tool for answering a specific kind of question: "how many roots can this equation have?"

Bounding the number of roots of a polynomial

Here is the key observation: between any two roots of f, there is at least one root of f'.

Why? If f(a) = 0 and f(b) = 0 with a < b, then f(a) = f(b), the function is continuous on [a, b] and differentiable on (a, b) (polynomials are always both), so Rolle's theorem says f'(c) = 0 for some c in (a, b).

This gives an immediate bound: if f' has n real roots, then f has at most n + 1 real roots. Because between every consecutive pair of roots of f, there must be a root of f', and those roots of f' are all distinct (they live in non-overlapping intervals). So f having k roots forces f' to have at least k - 1 roots. Flipping this: if f' has n roots, then k - 1 \le n, meaning k \le n + 1.

Showing an equation has at most one root

The same idea runs in reverse. If you can show that f'(x) is never zero in an interval, then f cannot have more than one root there.

Common confusions

• "Rolle's theorem says the derivative is zero at every point between a and b." No — it says at least one such point exists. The derivative can be nonzero at every other point in the interval.

• "The point c is always the midpoint." There is no reason to expect c to be at the midpoint (a+b)/2. For a symmetric function like f(x) = x^2 - 1 on [-1, 1], the zero of f' is at x = 0, which happens to be the midpoint. But for f(x) = x^3 - x on [-1, 1], the zeros of f' are at x = \pm 1/\sqrt{3}, neither of which is the midpoint.

• "If f'(c) = 0, then c is a root of f." No. f'(c) = 0 means the tangent is horizontal at c — it says nothing about whether f(c) is zero. The function value at c can be anything.

• "Rolle's theorem works for any function with f(a) = f(b)." Only if the function is also continuous on [a, b] and differentiable on (a, b). Drop either condition and the conclusion can fail, as the counterexamples above show.

• "The converse is true: if f'(c) = 0 for some c in (a, b), then f(a) = f(b)." False. Take f(x) = x^3 on [-1, 2]. Here f'(0) = 0, but f(-1) = -1 \neq 8 = f(2).

Going deeper

The logical chain: Rolle → Lagrange → Cauchy

Rolle's theorem looks like a special case — it only handles functions with equal endpoint values. But it is the seed from which the entire Mean Value Theorem family grows. The logical chain is:

1. Rolle's theorem: if f(a) = f(b), there is a c where f'(c) = 0.
2. Lagrange's Mean Value Theorem: for any continuous-differentiable function on [a, b], there is a c where f'(c) = \frac{f(b) - f(a)}{b - a}. (Proved by applying Rolle's theorem to a cleverly tilted version of f.)
3. Cauchy's Mean Value Theorem: for two functions f and g, there is a c where \frac{f'(c)}{g'(c)} = \frac{f(b) - f(a)}{g(b) - g(a)}. (Proved by applying Rolle's theorem to a combination of f and g.)

Each theorem is proved by reducing to the one before it. Rolle's theorem, the simplest case, does all the heavy lifting. You will see the details in the Mean Value Theorems article.

Rolle's theorem and the existence of roots of derivatives

Here is a deep consequence. Suppose a polynomial p(x) of degree n has n real roots (counting multiplicity). Then p'(x) has at least n - 1 real roots. This follows directly from Rolle's theorem applied to each consecutive pair of roots of p.

More generally, if p has k distinct real roots, then p' has at least k - 1 real roots. This observation is the starting point for the theory of root separation — understanding how the roots of a polynomial and its derivative interlace on the real line.

The connection to the Intermediate Value Theorem

Rolle's theorem uses the Extreme Value Theorem (a continuous function on [a, b] attains its max and min) but not the Intermediate Value Theorem directly. However, the two theorems are complementary: the IVT tells you that a continuous function crosses a value, while Rolle's theorem (and its descendants) tell you where the derivative is zero. Together, they form the two pillars of the qualitative theory of continuous functions.

Where this leads next

Rolle's theorem is the door to the entire Mean Value Theorem family, and through that family, to almost every result about the behaviour of differentiable functions.

• Mean Value Theorems — Lagrange's MVT and Cauchy's MVT, both proved using Rolle's theorem. The most important generalisation.
• Monotonicity — using the derivative to determine where a function increases or decreases. The proof relies directly on the MVT, which relies on Rolle.
• Maxima and Minima — First Derivative Test — finding peaks and valleys of functions. The critical-point idea (f'(c) = 0) is exactly the condition Rolle's theorem identifies.
• Maxima and Minima — Second Derivative Test — using the second derivative to classify critical points as maxima or minima.