Intermediate Value Theorem

The visual claim

Imagine a continuous curve drawn from (a, f(a)) to (b, f(b)) without lifting your pen. Pick any horizontal line y = y₀ whose height lies strictly between f(a) and f(b). The Intermediate Value Theorem (IVT) asserts that the curve must cross that horizontal line somewhere — there is at least one c ∈ (a, b) with f(c) = y₀.

Stated as a slogan: a continuous function on a closed interval never skips a value. If it starts below and ends above, it must equal everything in between, somewhere.

This sounds obvious. It feels obvious because the real numbers are constructed to make it true. The completeness axiom — every bounded set of reals has a least upper bound — is exactly the property that prevents continuous curves from sneaking past a horizontal line. On the rationals, where completeness fails, the theorem is false: f(x) = x² − 2 is continuous, takes the value −1 at x = 1 and the value +2 at x = 2, but never takes the value 0 anywhere on the rationals — because √2 is irrational.

The formal statement

Intermediate Value Theorem. Let f: [a, b] → ℝ be continuous on the closed interval [a, b]. For every value y strictly between f(a) and f(b), there exists at least one c ∈ (a, b) such that f(c) = y.

The hypothesis is just continuity — no differentiability, no monotonicity, no smoothness. The conclusion is existence — IVT does not say c is unique, nor does it tell you how to find it. You can have a wildly oscillating continuous function that takes the value y a thousand times; IVT only insists on at least one.

Proof sketch — by bisection

Assume without loss of generality that f(a) < y < f(b). Define g(x) = f(x) − y; we want a zero of g. Set a₀ = a, b₀ = b. At each step n:

1. Let m_n = (a_n + b_n)/2.
2. If g(m_n) ≤ 0, set a_n+1 = m_n, b_n+1 = b_n.
3. Otherwise set a_n+1 = a_n, b_n+1 = m_n.

The intervals nest, their lengths halve, and at every step g(a_n) ≤ 0 ≤ g(b_n). The endpoints converge to a common limit c by the nested intervals theorem (a consequence of completeness). Continuity passes the limit through g: g(c) = lim g(a_n) ≤ 0 ≤ lim g(b_n) = g(c), so g(c) = 0, meaning f(c) = y.

This proof is the bisection algorithm. Run it on a computer, and after n steps you have located c within (b − a)/2ⁿ. After 30 iterations on the unit interval, you have c to better than one part in a billion.

Worked example: x³ − x − 1 = 0 has a real root

Let f(x) = x³ − x − 1. Compute f(1) = 1 − 1 − 1 = −1 and f(2) = 8 − 2 − 1 = 5. The function is a polynomial, hence continuous on [1, 2]. Since 0 lies between −1 and 5, IVT guarantees at least one c ∈ (1, 2) with f(c) = 0 — a real root.

IVT does not give a closed-form value of c. (For this cubic, a closed form exists via Cardano's formula, but it's ugly.) What IVT delivers is the certainty of existence — and the bisection method delivers numerical convergence: c ≈ 1.32472. The same argument shows that every odd-degree polynomial with real coefficients has at least one real root.

IVT vs Bolzano vs Extreme Value Theorem

	Intermediate Value Theorem	Bolzano's theorem	Extreme Value Theorem
Hypothesis	f continuous on [a, b]	f continuous on [a, b], f(a)·f(b) < 0	f continuous on [a, b]
Conclusion	f takes every value between f(a), f(b)	f has a root in (a, b)	f attains its max and min on [a, b]
Existence of	Pre-image of any intermediate y	A zero crossing	Global extrema
Proof rests on	Completeness of ℝ	Special case of IVT	Heine-Borel (compactness)
Special case of?	—	IVT (with y = 0)	—
Founding mathematician	Cauchy/Bolzano (1817)	Bolzano (1817)	Weierstrass (1860s)
Constructive?	Yes (bisection)	Yes (bisection)	No (max may be limit point)

All three theorems are pillars of single-variable analysis. IVT and Bolzano are about codomain coverage; the Extreme Value Theorem is about attaining the supremum. Together they make a continuous function on a closed interval as well-behaved as one could ask: it hits every intermediate value, it attains its extremes, and it can be evaluated reliably anywhere.

Classical applications

Bisection method. The earliest reliable root-finder in numerical analysis. It always converges, with predictable rate (one bit per iteration), and only requires a sign change — no derivatives. Bisection is the IVT's algorithmic incarnation.

Existence of fixed points. If f: [a, b] → [a, b] is continuous, then g(x) = f(x) − x satisfies g(a) ≥ 0 and g(b) ≤ 0. IVT gives a fixed point — a one-dimensional Brouwer.

Pancake (ham-sandwich) theorems. Every closed plane region can be cut into two equal areas by some straight line at any chosen angle. The proof: as you slide a perpendicular sweep line, the area on one side changes continuously from 0 to the total; IVT pins down the half-area cut.

Existence of an antipode of equal temperature. Going around the equator of the Earth, surface temperature is a continuous function of position. IVT applied to f(x) = T(x) − T(antipode of x) — which switches sign — forces the existence of two opposite points at the same temperature.

Solvability of polynomial equations. Every odd-degree real polynomial has a real root: as x → ±∞ the leading term dominates with opposite signs, so by IVT the polynomial crosses zero somewhere.

When the hypothesis fails

Discontinuity destroys the conclusion. Define f(x) = 0 for x < 0 and f(x) = 1 for x ≥ 0. On [−1, 1], f(−1) = 0 and f(1) = 1, but no c satisfies f(c) = 0.5. The function jumps over the value.

Open intervals can fail. The function f(x) = 1/x on (0, 1] has no maximum (it diverges as x → 0⁺). IVT itself still holds on each closed sub-interval, but Extreme Value fails on the half-open one.

Discontinuous bijections. A function f: [0, 1] → [0, 1] can be a bijection without being continuous; if so, it can't be an IVT example because it would be a homeomorphism.

Common mistakes

• Demanding a unique c. IVT is a some-c theorem. If f oscillates, many values of c can satisfy f(c) = y. The theorem says nothing about which one.
• Asking IVT to give you c. It is an existence theorem. Use bisection (or Newton's method, when applicable) to compute c.
• Forgetting the closed interval. IVT requires continuity on [a, b], including the endpoints. Continuity only on (a, b) is not enough — endpoint values must be defined.
• Applying IVT in non-totally-ordered codomains. "Between" doesn't make sense for vectors or complex numbers. The theorem is specifically for real-valued continuous functions.
• Confusing IVT with surjectivity. IVT only says f hits every value between f(a) and f(b). It says nothing about values above f(b) or below f(a).
• Mistaking the converse. A function with the IVT property need not be continuous — Darboux's theorem produces continuous-like (Darboux) functions that are not continuous. IVT is a one-way implication.