The Intermediate Value Theorem

The IVT is the standard tool for proving polynomials of odd degree have a real root: for $p(x) = x^{2k+1} + \cdots$, one shows $p(x) \to -\infty$ as $x \to -\infty$ and $p(x) \to +\infty$ as $x \to +\infty$, then applies Bolzano's Theorem on a sufficiently large interval $[-N, N]$. In numerical analysis, the bisection method for root-finding is literally the constructive skeleton hidden inside the IVT's proof, repeatedly halving an interval known to contain a sign change; it underlies more refined algorithms (Newton's method, secant method) used throughout scientific computing. In economics and game theory, one-dimensional and multi-dimensional generalizations (Brouwer's fixed point theorem in $\mathbb{R}^n$, and its extension Kakutani's theorem) underpin existence proofs for market equilibria and Nash equilibria; the IVT's fixed-point corollary above is the simplest ancestor of that entire machinery. In topology, the IVT generalizes to the statement that continuous images of connected sets are connected; this is the modern lens through which the theorem is properly understood, and it explains why an interval (a connected domain) is essential while a union of two disjoint intervals is not — foreshadowing the deeper study of Continuity on Intervals and Sets and, more specifically, Monotone Functions and Their Limits, where sign-change and injectivity arguments closely related to IVT reasoning determine when continuous functions have continuous inverses. In cryptography and coding theory, sign-change arguments over real-valued auxiliary functions are used in some algorithmic constructions and error-bound estimates, though there the IVT typically appears as a lemma inside larger numerical or algebraic arguments rather than as headline machinery.

:::quiz[If $f$ is continuous on $[0,4]$ with $f(0) = 3$ and $f(4) = -1$, which value must $f$ attain somewhere in $(0,4)$?]

• $5$
• $-2$
• $0$ ✓
• $4$

:::quiz[In the proof, the set $S = \{x \in [a,b] : f(x) \le y\}$ is used to]

• directly evaluate $f$ at $y$
• produce a candidate crossing point via its supremum, using completeness ✓
• show $f$ is bounded on $[a,b]$
• prove $f$ is monotonic

:::quiz[For the fixed point corollary, why must $f$ map $[a,b]$ into $[a,b]$ (not just be continuous on $[a,b]$)?]

• So that $g(x) = f(x) - x$ is well-defined with $g(a) \ge 0$ and $g(b) \le 0$, enabling Bolzano's Theorem ✓
• So that $f$ is automatically differentiable
• So that $f$ is automatically injective
• It is not actually needed; any continuous $f:[a,b]\to\mathbb{R}$ has a fixed point in $[a,b]$

Ask the tutor: How does the proof of the IVT change if we only assume $f$ has one-sided limits at every point instead of full continuity?

Picture the graph of $f$ as a curve drawn on $[a,b]$ without lifting the pen — this "no lifting the pen" heuristic is exactly what continuity encodes rigorously. If the curve starts below the height $y$ (at $x=a$, since $f(a) < y$) and ends above the height $y$ (at $x = b$, since $f(b) > y$), then somewhere it must cross the horizontal line at height $y$. The pen cannot teleport from below the line to above it. The proof formalizes "somewhere it must cross" by identifying the first place the curve fails to stay below $y$: consider the set of all $x$ where $f$ is still $\le y$, and take its supremum. Completeness guarantees this supremum exists as a real number; continuity then pins down that $f$ equals $y$ exactly there — it cannot overshoot or undershoot, because either would create room for points on the wrong side of the supremum in violation of its defining property. ```

```

The proof technique is a supremum / least-upper-bound argument: build the "boundary set" $S$ using the hypothesis, invoke completeness to get a candidate crossing point, then use continuity twice (once to rule out overshoot, once to rule out undershoot) to pin the value exactly. No step invokes the IVT or any theorem built from it — the only external input is completeness and the $\varepsilon$–$\delta$ definition of continuity, so the proof is non-circular.

Example 1 — locating a root of a transcendental equation to a guaranteed interval. Let $f(x) = x^3 - x - \cos(\pi x/4)$ on $[1, 2]$. We claim $f$ has a root in $(1,2)$. Continuity. $x^3 - x$ is a polynomial, hence continuous everywhere. $\cos(\pi x /4)$ is a composition of the continuous functions $x \mapsto \pi x/4$ and $\cos$, hence continuous everywhere. The difference of two continuous functions is continuous, so $f$ is continuous on all of $\mathbb{R}$, in particular on $[1,2]$. Sign check at endpoints.

So $f(1) < 0 < f(2)$. Conclusion. $f$ is continuous on $[1,2]$ and changes sign, so by Bolzano's Theorem there exists $c \in (1,2)$ with $f(c) = 0$, i.e. $c^3 - c = \cos(\pi c/4)$. This is an existence statement only — the proof gives no formula for $c$ — but it is airtight: no algebraic solution of this transcendental equation is needed to know a solution exists in $(1,2)$. Refinement (bisection, using the same theorem repeatedly). Evaluate at the midpoint $x = 1.5$:

Since $f(1) < 0$ and $f(1.5) > 0$, the same corollary applied to $[1, 1.5]$ locates the root in $(1, 1.5)$. Iterating this bisection (each step another application of Bolzano's Theorem) traps $c$ in intervals of length $2^{-n}$, giving a numerical method — this is exactly how the existence proof becomes an algorithm.

Example 2 — a fixed point that is not obvious from a formula, using non-endpoint data. Let $f(x) = \dfrac{1}{2}\cos(x) + \dfrac{x}{4} + 1$ on the interval $[0, 3]$. We show $f$ maps $[0,3]$ into itself and hence has a fixed point in $[0,3]$, without solving for it explicitly. Continuity. $\cos(x)$ and the linear function $x/4$ are continuous everywhere; sums and scalar multiples of continuous functions are continuous, so $f$ is continuous on $\mathbb{R}$, in particular on $[0,3]$. Range check. For all real $x$, $-1 \le \cos(x) \le 1$, so

For $x \in [0,3]$, $0 \le x/4 \le 3/4$. Adding these bounds together with the constant $1$:

So for every $x \in [0,3]$, $f(x) \in [0.5,\, 2.25] \subset [0,3]$. Thus $f: [0,3] \to [0,3]$. Apply the fixed point corollary. $f$ is continuous and maps $[0,3]$ into itself, so by the 1-dimensional Brouwer corollary there exists $x^* \in [0,3]$ with $f(x^*) = x^*$. Sanity check via direct sign analysis (independent confirmation). Let $g(x) = f(x) - x$. Then

Since $g(0) > 0 > g(3)$ and $g$ is continuous (difference of continuous functions), Bolzano's Theorem confirms directly a root of $g$ in $(0,3)$, matching the fixed point conclusion and illustrating that the fixed-point corollary is really Bolzano's Theorem in disguise.