Skip to content
Tutorials Center
My learning

Topics

Lesson 22 of 37 · Continuity

Uniform Continuity

Understand uniform continuity, how it differs from ordinary continuity, why continuous functions on [a, b] are uniformly continuous, and how to test it numerically.

In ordinary continuity, the δ you choose may depend on the point a. Uniform continuity asks for one δ that works at every point of the domain at once. The difference seems small, but it decides whether integrals exist, whether limits of continuous functions behave, and whether numerical grids converge.

Definition

f : D → ℝ is uniformly continuous on D if

Text
for every ε > 0 there exists δ > 0 such that
for all x, y ∈ D with |x - y| < δ we have |f(x) - f(y)| < ε.

Compare with continuity at each point: there the quantifiers read "for all a, for all ε, exists δ". Here they read "for all ε, exists δ, for all x, y". Moving "exists δ" in front of "for all points" is the whole difference. Uniformly continuous functions are continuous, but the converse can fail.

A failure: x² on ℝ

Take xₙ = n and yₙ = n + 1/n. Then |xₙ - yₙ| = 1/n → 0, but

Text
|yₙ² - xₙ²| = |yₙ - xₙ|·(yₙ + xₙ) = (1/n)(2n + 1/n) = 2 + 1/n² ≥ 2.

So for ε = 1 no δ works: whatever δ > 0 you pick, for large n the pair xₙ, yₙ is δ-close but their values differ by at least 2. The function x² is continuous on ℝ but not uniformly continuous. The function gets steeper as you go right, and a fixed δ cannot keep up.

Python x squared is not uniformly continuous on the real line

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Another failure: 1/x on (0, 1)

Take xₙ = 1/n and yₙ = 1/(n+1). Their distance is 1/(n(n+1)) → 0, yet |1/xₙ - 1/yₙ| = |n - (n + 1)| = 1. So 1/x is continuous on (0, 1) but not uniformly continuous there. The trouble is the open end at 0, where the function blows up.

The main theorem

Theorem (Heine-Cantor). If f is continuous on a closed bounded interval [a, b], then f is uniformly continuous on [a, b].

Proof. Suppose not. Then for some ε₀ > 0, for each n there are xₙ, yₙ ∈ [a, b] with |xₙ - yₙ| < 1/n and |f(xₙ) - f(yₙ)| ≥ ε₀. By Bolzano-Weierstrass, a subsequence x_{nₖ} converges to some c ∈ [a, b]. Since |y_{nₖ} - x_{nₖ}| < 1/nₖ, also y_{nₖ} → c. By continuity at c, both f(x_{nₖ}) and f(y_{nₖ}) converge to f(c), so their difference tends to 0, contradicting that it is at least ε₀. ∎

This is why compactness (see the Extreme Value Theorem) matters: it turns a point-by-point property into a global one.

Useful sufficient conditions

  • Lipschitz functions. If |f(x) - f(y)| ≤ K|x - y| for all x, y then δ = ε / K works, so f is uniformly continuous. Example: sin on ℝ with K = 1, since |sin x - sin y| ≤ |x - y|.
  • Bounded derivative. If f is differentiable with |f'| ≤ K, the Mean Value Theorem (see that lesson) gives the Lipschitz bound.
  • Square root. √x is uniformly continuous on [0, ∞) even though its derivative is unbounded near 0. For x, y ≥ 0, |√x - √y| ≤ √|x - y|, so δ = ε² works.
Python Check the delta formulas for sin and sqrt

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Why it matters

Uniform continuity lets you control a function on a whole interval from a single mesh size. For example, to show a continuous function is Riemann integrable (see the lesson on integrable functions), you choose one δ, then any partition with mesh below δ makes every subinterval's oscillation smaller than ε. Without uniformity, the mesh needed could change from place to place.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. How does uniform continuity differ from continuity at every point?

    In uniform continuity delta depends only on epsilon, not on the location of the points.
  2. 2. Which function is continuous but not uniformly continuous on its domain?

    Points 1/n and 1/(n+1) get arbitrarily close, but the values of 1/x differ by exactly 1.
  3. 3. Heine-Cantor says a continuous function on which kind of domain is uniformly continuous?

    Compactness of [a, b] lets us pass to a convergent subsequence in the proof.
  4. 4. If |f(x) - f(y)| is at most K|x - y| for all x and y, which delta works for a given epsilon?

    With delta = epsilon / K, |x - y| < delta gives |f(x) - f(y)| < K * delta = epsilon.

Summary

  • Uniform continuity uses one δ for the whole domain; it implies continuity but not conversely.
  • x² on ℝ and 1/x on (0, 1) are continuous but not uniformly continuous.
  • Heine-Cantor: continuous on [a, b] implies uniformly continuous.
  • Lipschitz functions are uniformly continuous with δ = ε / K; √x works with δ = ε².

Next: The Derivative as a Limit