Real Analysis 25 | Uniform Convergence [dark version]

Q: How does the quantifier order distinguish pointwise from uniform convergence?

Pointwise convergence allows the index n to depend on the specific point x: for each x and ε > 0, there exists n(x,ε) such that for all k ≥ n(x,ε), |F_k(x) − f(x)| 0, there exists a single N such that for all k ≥ N and for all x ∈ I, |F_k(x) − f(x)| < ε. That single N must work across the entire interval.

Q: What does the “ε-tube” picture mean for graphs of F_n?

Fix ε and draw the region consisting of all points within vertical distance ε of the limit graph f. Uniform convergence means that after some stage N, every graph of F_n lies completely inside this tube for every x in I. Pointwise convergence would only guarantee that each fixed x eventually lands inside the tube, possibly at different times for different x.

Q: Why introduce the supremum norm, and how does it connect to uniform convergence?

To measure closeness globally, define ||f − g||_∞ = sup_{x∈I} |f(x) − g(x)|, capturing the worst-case pointwise error across I. Then uniform convergence is equivalent to the statement that the worst-case error between F_n and f goes to zero: ||F_n − f||_∞ → 0 as n → ∞. This turns uniform convergence into ordinary convergence of real numbers.

Q: What goes wrong in the jump-function example where pointwise convergence holds but uniform convergence fails?

The limit function f has a jump, with values −1 and 1 on either side (as given). Near the jump, the distance |F_n(x) − f(x)| does not shrink uniformly: the supremum norm stays at least 1. Since ||F_n − f||_∞ never approaches 0, uniform convergence fails even though each fixed x still sees F_n(x) → f(x).

Q: Which properties does uniform convergence preserve that pointwise convergence may not?

Uniform convergence is highlighted as preserving continuity and boundedness. If every F_n is continuous and F_n converges uniformly to f, then f remains continuous. Similarly, if each F_n is bounded and convergence is uniform, the limit function f is also bounded. Pointwise convergence alone does not guarantee these inheritances.

TL;DR

Uniform convergence requires a single N for each ε that works for all x in the domain, unlike pointwise convergence where N may depend on x.

Briefing Cornell Notes

Briefing

Uniform convergence is the stronger notion of convergence for functions where a single “eventually” index works for every point in the domain at once. In pointwise convergence, the index n allowed to make |F_n(x) − f(x)| small can depend on the specific point x. Uniform convergence swaps that flexibility: for every ε > 0, there is one N such that for all n ≥ N and for every x in the interval I, the inequality |F_n(x) − f(x)| < ε holds simultaneously. That quantifier order change is the whole difference—and it matters because it turns many separate point-by-point guarantees into one global control over the entire graph.

Geometrically, the limit function f has a “tube” of height ε around its graph. Uniform convergence means that from some stage onward, every graph of F_n lies entirely inside that tube for all x in I, not just at individual x-values. Pointwise convergence only ensures that each fixed x eventually lands in the tube, even if different points require different stages. The uniform version forces the entire family of functions to settle down together.

To make this precise, the transcript introduces a way to measure how close two functions are: use the supremum norm. Given two functions f and g on I, consider |f(x) − g(x)| for each x, then take the maximum “worst-case” size across the domain. Since a maximum may not exist, the supremum is used instead: ||f − g||_∞ = sup_{x∈I} |f(x) − g(x)|. With this distance, uniform convergence becomes a simple statement about ordinary convergence of real numbers: F_n → f uniformly on I exactly when ||F_n − f||_∞ → 0 as n → ∞. In other words, the worst-case error across the whole interval shrinks to zero.

An example illustrates why pointwise convergence alone is insufficient. The sequence consists of functions that get steeper and steeper, converging pointwise to a limit function f that has a jump. Even though each fixed x sees F_n(x) approach f(x), the supremum norm never drops below a positive threshold. The transcript notes that around the jump, the distance between F_n and f stays at least 1 (using the given values −1 and 1 on either side of the jump), so ||F_n − f||_∞ cannot tend to 0. This shows pointwise convergence does not imply uniform convergence.

The key takeaway is that uniform convergence is strictly stronger than pointwise convergence. That strength pays off later because uniform convergence preserves important properties of functions. The transcript highlights two: continuity and boundedness. While pointwise convergence can fail to keep these properties, uniform convergence ensures that if each F_n is continuous (or bounded) and F_n converges uniformly to f, then the limit function f inherits continuity (or boundedness).

Cornell Notes

Uniform convergence strengthens pointwise convergence by requiring one index N that works for every point x in the domain at the same time. Formally, for every ε > 0 there exists N such that for all n ≥ N and all x ∈ I, |F_n(x) − f(x)| < ε. This global control can be expressed using the supremum norm: ||F_n − f||_∞ = sup_{x∈I} |F_n(x) − f(x)|, and uniform convergence is equivalent to ||F_n − f||_∞ → 0. A jump-function example shows why pointwise convergence alone fails: the supremum error stays bounded away from zero, so the supremum norm cannot go to 0. Uniform convergence is therefore stronger and preserves properties like continuity and boundedness.

How does the quantifier order distinguish pointwise from uniform convergence?

Pointwise convergence allows the index n to depend on the specific point x: for each x and ε > 0, there exists n(x,ε) such that for all k ≥ n(x,ε), |F_k(x) − f(x)| < ε. Uniform convergence moves the “exists n” outside the “for all x”: for each ε > 0, there exists a single N such that for all k ≥ N and for all x ∈ I, |F_k(x) − f(x)| < ε. That single N must work across the entire interval.

What does the “ε-tube” picture mean for graphs of F_n?

Fix ε and draw the region consisting of all points within vertical distance ε of the limit graph f. Uniform convergence means that after some stage N, every graph of F_n lies completely inside this tube for every x in I. Pointwise convergence would only guarantee that each fixed x eventually lands inside the tube, possibly at different times for different x.

Why introduce the supremum norm, and how does it connect to uniform convergence?

To measure closeness globally, define ||f − g||_∞ = sup_{x∈I} |f(x) − g(x)|, capturing the worst-case pointwise error across I. Then uniform convergence is equivalent to the statement that the worst-case error between F_n and f goes to zero: ||F_n − f||_∞ → 0 as n → ∞. This turns uniform convergence into ordinary convergence of real numbers.

What goes wrong in the jump-function example where pointwise convergence holds but uniform convergence fails?

The limit function f has a jump, with values −1 and 1 on either side (as given). Near the jump, the distance |F_n(x) − f(x)| does not shrink uniformly: the supremum norm stays at least 1. Since ||F_n − f||_∞ never approaches 0, uniform convergence fails even though each fixed x still sees F_n(x) → f(x).

Which properties does uniform convergence preserve that pointwise convergence may not?