Skip to content

Heine-Cantor Theorem

Theorem

Let a<ba < b be real numbers, and consider a function f ⁣:[a,b]→Rf\colon [a, b] \to \mathbb{R} that is continuous on [a,b].[a, b]\text{.} Then, ff is uniformly continuous on [a,b].[a, b]\text{.}


Interpretation

Continuity of ff at each point gives, for every ε>0\varepsilon>0, a δ\delta that may depend on the point. The Heine-Cantor theorem says that on a closed bounded interval this dependence can be removed: one δ\delta works at every point of [a,b][a,b] at once. The function cannot become steeper and steeper towards an edge, because the endpoints belong to the interval and ff is continuous there as well.

The graph of x squared over 2 on the interval from a to b with three equal-sized rectangles of width 2 delta and height 2 epsilon centred on the graph, each containing the graph over its whole width.

The figure takes f(x)=12x2f(x)=\tfrac{1}{2}x^2 with a=0a=0 and b=2b=2. Each rectangle has width 2δ2\delta and height 2ε2\varepsilon and is centred on a point of the graph, and the graph stays inside every rectangle over its whole width. Since ∣f′(x)∣=∣x∣≤2|f'(x)|=|x|\le2 on [0,2][0,2], the mean value theorem gives ∣f(x1)−f(x2)∣≤2∣x1−x2∣|f(x_1)-f(x_2)|\le2|x_1-x_2|, so δ=ε/2\delta=\varepsilon/2 works everywhere.

Both properties of the domain matter. On the open interval (0,1)(0,1) the continuous function 1/x1/x is not uniformly continuous: it becomes arbitrarily steep near the missing endpoint 00. On the closed but unbounded interval [0,∞)[0,\infty) the function x2x^2 fails too: the points xx and x+δ2x+\tfrac{\delta}{2} are closer than δ\delta, yet their values differ by xδ+δ24x\delta+\tfrac{\delta^2}{4}, which exceeds any ε\varepsilon for large xx. The theorem gives a sufficient condition only, as the Examples show.

The theorem is the usual step in showing that continuous functions on [a,b][a,b] are Riemann integrable: one δ\delta controls the variation of ff on every subinterval of a partition at once.

"Uniform" does not mean that δ\delta is independent of ε\varepsilon. It still depends on ε\varepsilon, only no longer on the point.

The theorem returns later with [a,b][a,b] replaced by more general closed and bounded sets, the compact sets.


Key Results

  • This theorem has a generalized version if [a,b][a,b] is replaced by a compact set.
  • This theorem does not imply that uniform continuity holds only on closed intervals.

Examples

  • The function f(x)=x2+sin⁡xexf(x)=\dfrac{x^2+\sin x}{\mathrm{e}^x} is uniformly continuous on any closed interval since we know it is continuous everywhere.
  • The function g(x)=xg(x)=x is uniformly continuous on (0,1)(0,1) though (0,1)(0,1) is not a closed interval. This does not contradict the theorem since the "closed interval" condition is sufficient but not necessary.

Prerequisites

Sources

Symbols: delta, epsilon, infinity, real numbers

Last updated: 20 October 2026 03:02