Uniform convergence of function sequences: Foundations

This page assumes nothing. Before you can even read the parent note Uniform convergence, you must own every symbol it throws at you. We build each one from a picture.

1. Numbers we allow: $R$ and the interval $[a, b]$

Before we name anything else, fix the playground of numbers we work in.

Picture $R$ as the entire horizontal ruler stretching forever both ways, and $[a, b]$ as a highlighted stretch of it with solid dots at both ends.

WHY the topic needs this: every input we ever feed a function is a real number, and the domain we look at is almost always either all of $R$ or a segment $[a, b]$ . We must own these two symbols first, since they appear the moment we describe where a function lives.

2. What is a function $f$ ?

Picture it as a curve on a grid. The horizontal axis holds the input $x$ ; the vertical axis holds the output $f (x)$ . The whole curve is the function — one height for every horizontal position.

Figure — Uniform convergence of function sequences

WHY the topic needs this: we are going to have many curves at once, and compare their heights. So first be totally comfortable that "a function" = "one curve = a rule assigning one height per $x$ ".

3. The common domain $E$

Picture $E$ as the stretch of the horizontal axis we are allowed to look at. When we later write "for all $x \in E$ ", the symbol $\in$ means "belongs to", so " $x \in E$ " reads " $x$ is one of the allowed inputs".

WHY the topic needs this: uniform convergence is not a property of a formula alone — it depends on which inputs you allow. The parent note shows $x^{n}$ is uniform on $[0, a]$ but not on $[0, 1]$ : same formula, different $E$ , different answer. So $E$ must be a first-class symbol, named before anything is measured over it.

4. What does the subscript in $f_{n}$ mean?

Picture a flip-book: each page $n$ shows a slightly different curve. As you flip faster ( $n$ grows), the curves settle into one final shape. That final shape is the limit function $f$ (no subscript — it is the destination, not a member of the list).

WHY the topic needs this: uniform vs pointwise is a statement about this flip-book settling down, so we need the counting index $n$ before anything else.

5. Absolute value $∣ \cdot ∣$ — measuring a gap

Now the crucial use: $∣ f_{n} (x) - f (x) ∣$ is the vertical distance between curve number $n$ and the target curve, measured at a single input $x \in E$ . It is always $\geq 0$ (a distance can't be negative), and it is $0$ exactly when the two curves touch at that $x$ .

WHY the topic needs this: convergence is this gap shrinking. Every definition below is a sentence about $∣ f_{n} (x) - f (x) ∣$ getting small.

6. The error budget $ε$ (epsilon)

Picture a thin horizontal band of half-height $ε$ drawn around the target curve — a fuzzy ribbon from $f (x) - ε$ up to $f (x) + ε$ . Saying " $∣ f_{n} (x) - f (x) ∣ < ε$ " means: at that $x$ , curve $f_{n}$ lies inside the ribbon.

WHY the topic needs this: convergence is a challenge game. The challenger names a budget $ε$ ; we must eventually fit inside it. Smaller $ε$ = thinner ribbon = harder challenge.

7. The threshold $N$ and the arrow $⟹$

The symbol $⟹$ means "leads to / guarantees". So $n \geq N ⟹ ∣ f_{n} (x) - f (x) ∣ < ε$ reads in plain words: "as soon as we are at flip-book page $N$ or later, curve $f_{n}$ sits inside the ribbon at $x$ ."

WHY the topic needs this: the single difference between pointwise and uniform convergence is whether $N$ is allowed to depend on $x$ . You cannot see that difference until $N$ is crystal clear.

8. The quantifiers $\forall$ and $\exists$ (with their domains)

$\forall x \in E \exists N \in N$ : "for each allowed input $x$ , you can then find some page number $N$ " — the $N$ is chosen after seeing $x$ , so it may depend on $x$ . (Each runner gets their own deadline.)
$\exists N \in N \forall x \in E$ : "there is one page number $N$ that then works for every allowed input $x$ " — the $N$ is chosen before $x$ , so it cannot depend on $x$ . (One deadline for the whole crowd.)

WHY the topic needs this: the parent note's ONE punchline is "the only difference is quantifier order". This is the deepest symbol on the page — reread it until it clicks.

9. The fully quantified $ε$ – $N$ definition of a numerical limit

Before we handle a limit of functions, nail the limit of ordinary numbers — because uniform convergence is built by re-wrapping this exact template.

WHY the topic needs this: the parent note's $M_{n} \to 0$ is a numerical limit; and continuity, integrals and Cauchy tests all reduce to statements of this shape. Owning the quantifier order here means you can read every $\to$ on the parent page precisely.

10. Supremum $sup$ — the tightest ceiling (and when it is finite)

First, one more symbol: throughout this section $g$ stands for a generic function on $E$ , i.e. a rule $g$ that takes any input $x \in E$ and returns a real number $g (x) \in R$ (exactly the "machine" of §2, just given a temporary name so we can talk about its collection of output values). In a moment we take $g (x) = ∣ f_{n} (x) - f (x) ∣$ , the gap.

Why not just say "maximum"? Because sometimes there is no maximum. Example: the values $x^{n}$ for $x \in [0, 1)$ climb toward $1$ but never equal $1$ (since $x < 1$ ). There is no largest value — yet $1$ is the tightest ceiling. So $sup = 1$ even though the max does not exist. This exact subtlety is what makes the $x^{n}$ counterexample work.

Now the star quantity of the whole topic — take $g (x) = ∣ f_{n} (x) - f (x) ∣$ : $M_{n} := sup_{x \in E} ∣ f_{n} (x) - f (x) ∣.$ In words: the biggest gap anywhere over $E$ between curve $n$ and the target. Picture sliding along the whole domain $E$ and recording the worst (tallest) vertical gap — that worst gap is $M_{n}$ .

WHY the topic needs this: $sup$ is the tool that converts a statement about infinitely many points into one checkable number — but only when that number is finite, so boundedness is part of the equipment.

11. The limit arrow $\to$ applied to $M_{n}$

We already gave the fully quantified meaning of $a_{n} \to L$ in §9. Apply it to the specific sequence $a_{n} = M_{n}$ and target $L = 0$ : $M_{n} \to 0 means \forall ε > 0 \exists N \in N \forall n \geq N : M_{n} < ε .$ So $M_{n} \to 0$ means the worst gap over $E$ shrinks to nothing. That single line is uniform convergence (parent §2). Separately, $x^{n} \to 0$ for a fixed $x \in [0, 1)$ means: multiplying a number below $1$ by itself again and again crushes it toward $0$ .

WHY the topic needs this: every convergence claim ends in " $\to$ something". It is the verb of the whole subject — and now you can read it as a precise chain of quantifiers.

12. Continuous and integrable

WHY the topic needs these: the payoff theorems (parent §4) say uniform convergence preserves continuity and lets you swap limit with integral. You must recognise these words to see why uniform convergence is "worth the trouble".

Prerequisite map

Read top to bottom: the real numbers and interval fix the playground; pictures of functions on a common domain $E$ feed the gap, epsilon-ribbon and $N$ -deadline, which feed the quantifier duel and the $ε$ – $N$ template for limits; the finite supremum turns "for all $x$ " into a single number whose limit defines uniform convergence, which unlocks the payoff theorems.

Equipment checklist

Recall Self-test — can you answer each before reading the parent note?

What do $R$ and $[a, b]$ mean? ::: $R$ = every point on the number line; $[a, b]$ = all reals from $a$ to $b$ including both endpoints. What does the letter $E$ stand for, and why can't it be ignored? ::: The common domain of inputs shared by all $f_{n}$ and $f$ ; uniformity depends on which inputs you allow ( $x^{n}$ is uniform on $[0, a]$ but not on $[0, 1]$ ). What does the subscript $n$ in $f_{n}$ mean? ::: A name-tag counting which curve in the list; it is NOT multiplication by $n$ . What does $∣ f_{n} (x) - f (x) ∣$ measure, geometrically? ::: The vertical distance between curve $n$ and the target curve at the single input $x \in E$ . What role does $ε$ play? ::: A chosen positive error budget; it draws a ribbon of half-height $ε$ around the target that $f_{n}$ must eventually sit inside. What does $n \geq N ⟹ \dots$ promise? ::: From page $N$ onward, the stated smallness is guaranteed; $N$ is the deadline. Write the fully quantified meaning of $a_{n} \to L$ . ::: $\forall ε > 0 \exists N \in N \forall n \geq N : ∣ a_{n} - L ∣ < ε$ . Why must you write $\forall x \in E$ and $\exists N \in N$ rather than bare $\forall x, \exists N$ ? ::: To say where each variable lives; and because $\forall x \exists N$ (N may depend on x) differs from $\exists N \forall x$ (one N for all x) — pointwise vs uniform. In $sup_{x \in E} g (x)$ , what is $g$ and why $sup$ not "maximum"? ::: $g$ is a generic function on $E$ (maps each $x \in E$ to a real $g (x)$ ); $sup$ is used because the biggest value may never be reached (e.g. $x^{n}$ on $[0, 1)$ approaches but never hits $1$ ), yet $sup$ still names the tightest ceiling — provided the gap is bounded on $E$ , else $M_{n} = + \infty$ . What is $M_{n}$ in one phrase, and what does $M_{n} \to 0$ mean? ::: $M_{n}$ = the worst (finite) gap over $E$ between $f_{n}$ and $f$ ; $M_{n} \to 0$ means that worst gap shrinks to nothing — this IS uniform convergence. What does uniform convergence let you safely do that pointwise cannot? ::: Keep continuity in the limit and swap $lim$ with $\int$ .

Connections

Parent: Uniform convergence — go here once every symbol above feels obvious.
Pointwise convergence — the weaker cousin these foundations also define.
Weierstrass M-test and Cauchy sequences in metric spaces — where $sup$ and $N$ reappear.
Continuity preserved under uniform limits and Interchange of limit and integral — the payoff vocabulary of §12.

#flashcards/maths