Measure and integration are where real analysis becomes a coherent system rather than a bag of clever \epsilon tricks. The definitions are chosen so that the theorems you want to be true actually become true, while the counterexamples tell you what cannot be demanded.

This article is a guided tour through measure and integration using worked examples. The theme is not to cover everything, but to show how the central moves repeat: build measurable sets, approximate complicated objects by simple ones, and use convergence theorems that have precise hypotheses.

Worked example: indicator functions and the meaning of an integral

For a measurable set E in [0,1], the indicator function 1_E is defined by 1_E(x)=1 if x∈E and 1_E(x)=0 otherwise.

A core fact is:

∫_0^1 1_E(x) dx = μ(E),

where μ is Lebesgue measure, which agrees with length on intervals.

This single identity is the bridge between sets and functions. Many problems about integrals become problems about the sizes of level sets.

Worked example: building measurable sets from open intervals

A standard measurable collection on ℝ is the Borel sets: start with open intervals and close under countable unions, countable intersections, and complements.

Example pattern: if E_n are open sets, then ⋃_n E_n is open, and its complement is closed, hence Borel. If F_n are closed sets, then ⋂_n F_n is closed, hence Borel. By iterating these closures you build extremely complicated sets while keeping measurability automatic.

This is a key habit: you do not prove measurability by explicit formulas; you prove it by closure properties.

Worked example: simple functions approximate measurable functions

A simple function is a finite linear combination of indicators:

s(x)=∑_{j=1}^m a_j 1_{E_j}(x),

where E_j are measurable and a_j are real numbers.

The Lebesgue integral is built by first defining the integral of nonnegative simple functions as

∫ s dμ = ∑_{j=1}^m a_j μ(E_j),

then defining the integral of a nonnegative measurable f as the supremum of ∫ s over simple s with 0 ≤ s ≤ f, and then extending by linearity to integrable functions.

A concrete approximation you can always keep in mind is range binning. Let f:[0,1]→[0,1] be measurable. For each n, partition [0,1] into 2^n bins of width 2^{-n} and define

s_n(x) = ∑_{k=0}^{2^n−1} (k/2^n) · 1_{ {x : k/2^n ≤ f(x) < (k+1)/2^n} }(x).

Then s_n is simple, 0 ≤ s_n ≤ f, and s_n(x) increases pointwise \to f(x) as n increases. This is not an ad hoc trick. It is one reason the integral behaves well under monotone limits.

Worked example: monotone convergence in action

Let f_n(x)=1_{[0,1−1/n]}(x) on [0,1]. Then f_n(x) increases pointwise \to 1_{[0,1)}(x). By the monotone convergence theorem,

∫_0^1 f_n(x) dx → ∫_0^1 1_{[0,1)}(x) dx.

Compute directly:

∫_0^1 f_n(x) dx = 1 − 1/n → 1,

and

∫_0^1 1_{[0,1)}(x) dx = 1.

The theorem matches calculation because the integral is defined in a way that respects increasing approximation by simple functions.

Worked example: dominated convergence and why it needs domination

Consider f_n(x)=sin(nx)/n on [0,2π]. Pointwise, f_n(x)→0. Also |f_n(x)| ≤ 1/n ≤ 1 for all x and all n, so the constant function g(x)=1 dominates every f_n and is integrable. Dominated convergence gives

∫_0^{2π} f_n(x) dx → 0.

You can compute the integral directly:

∫_0^{2π} sin(nx)/n dx = [−cos(nx)/n^2]_0^{2π} = 0.

So the theorem matches calculation. The pattern is what matters: oscillation is harmless when amplitude is uniformly controlled by an integrable bound.

In contrast, spike sequences show that pointwise convergence without any integrable domination does not allow exchanging limit and integral.

Worked example: L^p norms show different kinds of control

On [0,1], define h_n(x)=√n · 1_{[0,1/n]}(x). Then

||h_n||_1 = ∫_0^1 |h_n| dx = √n · (1/n) = 1/√n → 0,

but

||h_n||_2^2 = ∫_0^1 |h_n|^2 dx = n · (1/n) = 1,

so ||h_n||_2 = 1 for all n, and

||h_n||_∞ = √n → ∞.

This one example teaches three distinct notions of convergence:

• h_n → 0 in L^1
• h_n does not approach 0 in L^2
• h_n does not approach 0 uniformly

Different norms control different operations. If your argument needs worst-case error control, L^1 is not enough. If your argument needs quadratic energy control, L^2 is the right scale. If you need pointwise stability across the domain, the sup norm matters.

Worked example: differentiating under the integral sign as a limit exchange

A standard real-analysis question is when you can pass a derivative inside an integral:

d/dθ ∫_a^b F(x,θ) dx = ∫_a^b ∂_θ F(x,θ) dx.

A practical sufficient condition is:

• F(·,θ) is integrable for each θ
• ∂_θ F(x,θ) exists for almost every x
• and |∂_θ F(x,θ)| ≤ g(x) for an integrable g, uniformly for θ in a neighborhood

Then dominated convergence applied to the difference quotient gives the result.

Concrete example: I(θ)=∫_0^1 x^θ dx for θ > −1.

Direct computation gives I(θ)=1/(θ+1) and I'(θ)=−1/(θ+1)^2.

Inside the integral, ∂_θ x^θ = x^θ ln x, so the candidate is

∫_0^1 x^θ ln x dx.

Justifying the interchange is a domination problem: you need an integrable bound for |x^θ ln x| that holds uniformly for θ in a compact interval. This is a standard estimate, and one clean route uses the substitution x=e^{−t} \to turn the integral into a convergent integral on [0,∞) with an exponential weight. The key lesson is structural: interchanging derivative and integral is not symbolic; it is a limit exchange argument, and limit exchange is justified by uniform domination.

Worked example: almost everywhere is a feature, not a loophole

Many theorems conclude that something holds almost everywhere, meaning outside a set of measure zero. This is not a concession. It is a recognition that measure-zero exceptions do not affect integrals and L^p norms.

Example: define f(x)=1 on rational numbers in [0,1] and f(x)=0 on irrational numbers in [0,1]. The set of rationals has measure zero, so ∫_0^1 f(x) dx = 0. Yet f is discontinuous everywhere. This shows that measurability and integrability do not enforce pointwise regularity. Real analysis separates questions:

• If you care about integrals and averages, measure theory is the right tool.
• If you care about pointwise continuity, you need additional hypotheses.

The thread that ties the examples together

Every example above is a different face of one idea: the right notion of approximation depends on the operation you want to control.

• To control integrals, you need monotonicity, domination, or L^1 convergence.
• To control pointwise structure like continuity, you often need uniform convergence or equicontinuity.
• To control derivatives, you need uniform control on derivatives or domination applied to difference quotients.
• To control worst-case error, you need sup-norm bounds.

Measure and integration give real analysis its backbone because they provide a stable way to pass to limits, but they do so only under hypotheses that prevent mass from hiding in places your chosen notion of convergence cannot see.

If you work these examples until you can reproduce the core estimates without looking, you will have a practical command of the subject: not as a list of results, but as a disciplined method for deciding what is true, why it is true, and which hidden assumption would be required if it is not.

Worked example: Riemann integrable versus Lebesgue integrable

On a bounded interval, every Riemann integrable function is Lebesgue integrable and the integrals agree, but the Lebesgue integral handles limits more robustly. A simple illustration is the sequence of step functions that approximate a measurable function from below as in the range-binning construction. The point is not that step functions are special, but that the approximation can be arranged to be monotone, and monotonicity unlocks monotone convergence.

A practical habit is: if you can build an increasing sequence of simple functions s_n with s_n ↑ f, then you can compute or estimate ∫ f by computing ∫ s_n and taking a limit. This method naturally respects sets of measure zero, so you do not need to track behavior on negligible exceptional sets.

Worked example: convergence in measure is weaker than L^1 but still useful

A sequence f_n converges \to f in measure on [0,1] if for every ε>0,

μ({x : |f_n(x) − f(x)| > ε}) → 0.

This captures the idea that large errors occur on sets whose measure becomes small, but it does not force the average size of the error to vanish. Spikes again clarify the distinction.

Take the spike family f̃_n from earlier with integral 1 and pointwise limit 0. For any fixed ε>0, the set where f̃_n(x) > ε is essentially (0,1/n], whose measure is 1/n. So f̃_n → 0 in measure. Yet ∫ f̃_n = 1 for all n, so there is no convergence of integrals.

This shows why dominated convergence asks for domination and not merely convergence in measure: convergence in measure controls where the spikes are, but not how tall they can be.

Worked example: Fubini on a rectangle with a simple function

Let E be a measurable \subset of [0,1] and consider the function on the unit square [0,1]×[0,1],

F(x,y)=1_E(x).

This function does not depend on y. Its integral over the square is

∫_0^1 ∫_0^1 1_E(x) dy dx.

Compute the inner integral first: ∫_0^1 1_E(x) dy = 1_E(x) because the inner integral is over y and the integrand is constant in y. Therefore,

∫_0^1 ∫_0^1 1_E(x) dy dx = ∫_0^1 1_E(x) dx = μ(E).

If you reverse the order, you get the same result:

∫_0^1 ∫_0^1 1_E(x) dx dy = ∫_0^1 μ(E) dy = μ(E).

This is a toy computation, but it encodes the idea behind Fubini: under appropriate integrability hypotheses, iterated integrals agree with the integral over the product domain. In more advanced problems, the role of the toy calculation is to remind you what is being claimed, and which step uses integrability rather than algebra.

Worked example: absolute continuity as the integration-friendly notion of regularity

In the classical setting, a differentiable function with integrable derivative satisfies

f(b) − f(a) = ∫_a^b f'(x) dx.

In Lebesgue theory, the natural hypothesis for this identity is absolute continuity. An absolutely continuous function can be reconstructed from its derivative almost everywhere, and the derivative is integrable. This is one of the places where almost everywhere language is not a weakness but the correct interface between pointwise change and integral control.

The practical takeaway is: when you want to interchange differentiation and integration or recover a function from its derivative, the right regularity class is not merely continuous and not merely differentiable at many points, but absolutely continuous, because that notion is built to behave well under integration.

Real analysis is full of statements that look alike until you read the quantifiers. The fastest way to build proof skill is to stop asking what theorem applies and start asking what structure is available. Uniform convergence is a perfect place to practice this, because it sits at the intersection of pointwise limits, continuity, integration, differentiation, and completeness.

This guide is not a list of theorems to memorize. It is a set of proof strategies you can reuse, starting from the core definition and working outward to the standard stability results.

The definition you must be able to use on demand

Let (f_n) be a sequence of functions on a set E, and let f be a function on E. We say f_n → f uniformly on E if:

For every ε>0 there exists N such that for all n≥N and all x∈E,

|f_n(x) − f(x)| < ε.

The defining difference from pointwise convergence is the position of for all x∈E. Uniform convergence is a single-index guarantee over the whole domain.

A compact equivalent form is:

sup_{x∈E} |f_n(x) − f(x)| → 0.

Almost every proof about uniform convergence is a refined way of using that supremum control.

Strategy: build a uniform bound by factoring the expression

Many uniform convergence problems reduce to an estimate of the form

|f_n(x) − f(x)| ≤ A_n · B(x),

where B(x) is bounded on E and A_n → 0. Then the supremum is controlled:

sup_{x∈E} |f_n(x) − f(x)| ≤ A_n · sup_{x∈E} |B(x)| → 0.

Worked example: f_n(x) = x/(n+x) on [0,1].

The limit is f(x)=0. Compute

|f_n(x)| = x/(n+x) ≤ 1/n.

So sup_{x∈[0,1]} |f_n(x)| ≤ 1/n → 0. Uniform convergence is immediate.

This strategy is the default when the dependence on x can be bounded without losing the decay in n.

Strategy: use the Cauchy criterion for uniform convergence

Uniform convergence is convergence in the sup norm. That means it has a Cauchy characterization:

f_n converges uniformly on E if and only if for every ε>0 there exists N such that for all m,n≥N,

sup_{x∈E} |f_n(x) − f_m(x)| < ε.

This is especially useful when you do not know the limit f in advance. You can prove uniform convergence first, then define f as a pointwise limit and use completeness results to justify existence in an appropriate function space.

A common setting is a series of functions ∑_{k=1}^∞ u_k(x). You look at partial sums s_n(x)=∑_{k=1}^n u_k(x) and check whether (s_n) is uniformly Cauchy.

Strategy: apply the Weierstrass M-test for series

For a series ∑ u_k(x) on E, if you can find nonnegative numbers M_k such that

|u_k(x)| ≤ M_k for all x∈E

and ∑ M_k converges, then ∑ u_k converges uniformly.

The proof is a clean sup-norm tail estimate:

sup_{x∈E} |∑_{k=n+1}^m u_k(x)| ≤ ∑_{k=n+1}^m sup_{x∈E} |u_k(x)| ≤ ∑_{k=n+1}^m M_k,

and the tail ∑_{k=n+1}^∞ M_k goes \to 0.

Worked example: ∑_{k=1}^∞ sin(kx)/k^2 on ℝ.

We have |sin(kx)| ≤ 1, so |sin(kx)/k^2| ≤ 1/k^2. Since ∑ 1/k^2 converges, the M-test gives uniform convergence on all of ℝ. From this you can deduce continuity of the sum, and with more work you can justify differentiating term-by-term under stronger hypotheses.

The strategy lesson: if you can dominate every term by something summable that does not depend on x, you get uniform convergence essentially for free.

Strategy: localize difficulty using compact subintervals

On compact sets, continuous functions are bounded, and many families behave uniformly. Sometimes the right move is to restrict attention \to a compact set K and prove uniform convergence on K, then analyze what happens as K expands.

A standard pattern is: prove uniform convergence on [a,1] for every a>0, but not on (0,1]. The obstruction is typically a singularity near 0.

The disciplined move is:

• Fix a>0.
• On [a,1], use x≥a \to build clean bounds.
• Conclude uniform convergence on [a,1].
• Show failure on (0,1] by constructing x_n → 0 that keeps the error large.

This teaches you to separate global quantifiers from localized obstructions.

Strategy: prove stability results by \epsilon splitting

Uniform convergence preserves structure because you can choose one fixed index N that works everywhere. The standard proofs all look alike: bound an error by splitting it into three terms.

Continuity is preserved under uniform limits

Suppose each f_n is continuous on E and f_n → f uniformly. Fix x∈E and ε>0. Choose N such that |f_N(y) − f(y)| < ε/3 for all y∈E. Since f_N is continuous at x, choose δ such that |y−x|<δ implies |f_N(y) − f_N(x)| < ε/3. Then for |y−x|<δ,

|f(y) − f(x)| ≤ |f(y) − f_N(y)| + |f_N(y) − f_N(x)| + |f_N(x) − f(x)| < ε.

This is a template: pick one N using uniform convergence, then use the structure of f_N.

Integrals are preserved

If f_n → f uniformly on [a,b] and each f_n is integrable, then

∫_a^b f_n → ∫_a^b f.

Proof is a direct bound:

|∫_a^b f_n − ∫_a^b f| ≤ ∫_a^b |f_n − f| ≤ (b−a) sup_{x∈[a,b]} |f_n(x) − f(x)|.

Again, the sup norm is the control knob.

Derivatives can be exchanged under a strengthened hypothesis

A common theorem: if f_n are differentiable on [a,b], the derivatives f_n' converge uniformly \to a function g, and f_n(x_0) converges for some x_0 in [a,b], then f_n converges uniformly \to a differentiable function f with f'=g.

The proof strategy is to integrate derivative convergence:

f_n(x) − f_m(x) = (f_n(x_0) − f_m(x_0)) + ∫_{x_0}^x (f_n'(t) − f_m'(t)) dt.

Taking absolute values and supremums turns uniform convergence of derivatives plus convergence at the basepoint into a uniform Cauchy property for the functions.

Strategy: detect non-uniformity by constructing a witness sequence x_n

To prove uniform convergence fails, you show:

There exists ε>0 such that for every N there exist n≥N and x∈E with |f_n(x) − f(x)| ≥ ε.

A clean way is to build a specific x_n with persistent error.

Worked example: f_n(x)=x^n on [0,1]. The pointwise limit is f(x)=0 for 0≤x<1 and f(1)=1. Choose x_n = 1 − 1/n. Then f(x_n)=0, but

f_n(x_n) = (1 − 1/n)^n,

which stays bounded away from 0. Therefore |f_n(x_n) − f(x_n)| is not small for large n, so uniform convergence fails.

The strategy lesson: non-uniformity often concentrates near boundary points or near points where the limit function jumps.

Strategy: choose the right function space viewpoint

Uniform convergence is convergence in the sup norm: ||f_n − f||_∞ → 0. Once you frame problems in terms of norms, many arguments become one-line inequalities.

Examples:

• If ||f_n − f||_∞ → 0 on a finite-measure set, then ||f_n − f||_p → 0 for every p≥1 because ||h||_p ≤ μ(E)^{1/p} ||h||_∞.
• But ||f_n − f||_p → 0 does not imply uniform convergence; spike examples show how mass can concentrate.

The function space viewpoint keeps the implication structure honest.

A compact proof plan you can reuse

When you are given a convergence claim and asked to prove a structural property of the limit:

• Identify the contract: continuity, integrability, differentiability, interchange of limit, or preservation of bounds.
• Decide whether you need uniform convergence or something weaker such as domination or L^p convergence.
• If uniform convergence is plausible, try a uniform bound or an M-test first.
• If you cannot produce uniform bounds, look for a witness x_n that forces large error and then switch \to a weaker convergence notion that matches what is actually true.

Uniform convergence is not better than pointwise convergence in a moral sense. It is stronger, and it buys you precisely the right to pass limits through operations that are sensitive to worst-case behavior. Learning to prove and disprove uniform convergence is learning to control those worst-case behaviors directly.

Strategy: uniform convergence and exchanging limits with suprema

Some operations are even more sensitive than integrals. The supremum functional h ↦ sup_{x∈E} h(x) depends on the single worst point. Uniform convergence is designed to interact cleanly with it.

If f_n → f uniformly on E, then

| sup_{x∈E} f_n(x) − sup_{x∈E} f(x) | ≤ sup_{x∈E} |f_n(x) − f(x)|.

This inequality is worth memorizing because it is a model of how uniform convergence works: if you can bound an operation by the sup-norm error, then uniform convergence lets you pass the limit through that operation.

A common use is to justify minimizing and maximizing procedures. For instance, if each f_n is continuous on a compact set K and f_n → f uniformly, then the maxima of f_n converge to the maximum of f. The proof is a short application of the inequality above together with the fact that continuous functions achieve maxima on compact sets.

Strategy: control continuity uniformly via equicontinuity

Uniform convergence preserves continuity, but many problems give you only pointwise convergence and additional regularity information. In that situation, a strong tool is equicontinuity.

A family F of functions on K is equicontinuous if for every ε>0 there exists δ>0 such that for every f in F, and for all x,y in K with |x−y|<δ, we have |f(x)−f(y)|<ε.

The key point is that δ does not depend on the particular function. That is the same kind of uniformity that appears in uniform convergence.

One reason equicontinuity matters is the Arzelà–Ascoli theorem: on a compact K, any sequence of uniformly bounded and equicontinuous functions has a uniformly convergent subsequence. Even when you do not use the full theorem, the mindset helps: if you can prove a uniform modulus of continuity for all f_n, you are already halfway to uniform convergence results, because the behavior cannot concentrate into sharper and sharper spikes.

Worked example: a uniformly convergent geometric series of functions

Let 0<r<1 and define u_k(x)=r^k cos(kx) on ℝ. Since |cos(kx)|≤1, we have |u_k(x)| ≤ r^k for all x. The numerical series ∑ r^k converges, so the M-test gives that ∑ u_k converges uniformly on ℝ.

This single line proof gives you several consequences at once:

• The sum function is continuous because it is a uniform limit of continuous functions.
• You can integrate term-by-term on any finite interval because uniform convergence controls the integral error.
• You can often justify differentiating term-by-term if you strengthen the bound to control the derivatives as well.

The lesson: when a family has a built-in geometric decay factor that does not depend on x, uniform convergence is usually the right notion and the right theorems apply cleanly.

Real analysis becomes clear when you stop treating definitions as ceremonial and start treating them as contracts. A definition tells you what you are allowed to use, what you must prove, and what the statement is not saying. Counterexamples are the quickest way to learn those contracts because they expose the hidden clause you were unconsciously assuming.

This article develops one family of counterexamples that repeatedly shows up across limits, continuity, integration, and functional convergence. It is not a weird pathology. It is a disciplined construction that forces you to separate three ideas that look similar early on:

• Pointwise convergence versus uniform convergence
• Convergence of functions versus convergence of integrals
• Close most of the time versus controlled everywhere

The main character is a sequence of functions that forms a needle: tall, narrow spikes that move around. Your intuition says a spike that gets narrower should go away. Real analysis replies: only if you can control it in the correct sense.

The spike sequence

Define a sequence of functions on the interval [0,1] by

f_n(x)= n if 0≤x≤1/n, and f_n(x)=0 if 1/n < x ≤ 1.

You can picture f_n as a rectangle: height n, width 1/n, sitting at the left end of the interval. Its area is exactly

∫_0^1 f_n(x) dx = n · (1/n) = 1.

So every f_n has integral equal \to 1.

Now ask: what is the pointwise limit of f_n as n→∞?

Fix a point x in (0,1]. For sufficiently large n, we have 1/n < x. That means for large n, x lies in the region where f_n(x)=0. Therefore,

lim_{n→∞} f_n(x) = 0 for every x in (0,1].

At the single point x=0, the value is f_n(0)=n, which diverges. So the pointwise limit exists for all x in (0,1] and equals 0, while at 0 the sequence diverges.

If you modify the definition by setting f_n(0)=0 for all n, you obtain a sequence that converges pointwise \to 0 on all of [0,1]. For the conceptual lesson, it is better to keep the discussion on (0,1] or to redefine at 0 so pointwise convergence holds everywhere. We will use the modified version:

f̃_n(x)= n if 0<x≤1/n, f̃_n(x)=0 if 1/n<x≤1, and f̃_n(0)=0.

Then f̃_n → 0 pointwise on [0,1], and ∫_0^1 f̃_n = 1 still holds.

This is already the first counterexample:

• f̃_n → 0 pointwise
• but ∫ f̃_n does not approach ∫ 0 because 1 does not approach 0

So pointwise convergence does not justify exchanging limit and integral.

The surprising part is that nothing mystical occurred. The functions are simple. The failure comes from the absence of a uniform control that would let you pass the limit through the integral sign.

Where the proof attempt breaks

Many learners try to prove

lim_{n→∞} ∫_0^1 f̃_n(x) dx = ∫_0^1 lim_{n→∞} f̃_n(x) dx

by reasoning like this:

Since f̃_n(x) → 0 for each x, the values should eventually be small, so their integrals should eventually be small.

The hidden assumption is a uniformity assumption: you are acting as though there exists an index N such that for all n≥N and for all x in [0,1], the value |f̃_n(x)| is small. That is exactly uniform convergence.

But uniform convergence fails dramatically here. In fact,

sup_{x∈[0,1]} |f̃_n(x) − 0| = sup_{x∈[0,1]} f̃_n(x) = n,

so the supremum grows, not shrinks.

Pointwise convergence tells you: for each fixed x, you can make f̃_n(x) small by taking n large enough, but that index depends on x. There is no single N that works for all points at once.

The spike uses that gap. Every time you increase n, the spike gets narrower, but it also gets taller, and the supremum control gets worse.

A second lesson: pointwise limits of continuous functions can be discontinuous

Now consider the antiderivative (cumulative integral) of f̃_n:

F_n(x) = ∫_0^x f̃_n(t) dt.

We can compute F_n explicitly.

• If 0 ≤ x ≤ 1/n, then F_n(x) = ∫_0^x n dt = n x.
• If x > 1/n, then F_n(x) = ∫_0^{1/n} n dt = 1.

So

F_n(x) = n x for 0 ≤ x ≤ 1/n, and F_n(x)=1 for 1/n < x ≤ 1.

Each F_n is continuous, and the family F_n converges pointwise \to a discontinuous function:

F(0)=0 and F(x)=1 for 0<x≤1.

This is another core counterexample: a pointwise limit of continuous functions need not be continuous.

Again, the reason is that pointwise convergence is too weak. Uniform convergence would preserve continuity, but we do not have it. Near 0 the behavior changes with n.

This construction is an entry point to three ideas that should stay separate:

• Continuity is about local behavior at each point
• Uniform continuity is about a single modulus working everywhere
• Uniform convergence is about a single index controlling error everywhere

The spike sequence shows how local control can fail to assemble into global control.

The same idea, moved around: traveling spikes

The simplest spike sits at 0, but you can move it. Let x_n be points in [0,1], and define

g_n(x) = n if |x − x_n| ≤ 1/(2n), and g_n(x)=0 otherwise.

Then each g_n has integral 1 on [0,1] as long as the spike interval stays inside the domain (or you adjust at the boundary). For each fixed x, if x_n avoids x eventually, then g_n(x) → 0. If x_n hits x infinitely often, then g_n(x) does not converge.

This allows you to craft examples for any pattern you need:

• Spikes that converge pointwise almost everywhere but not everywhere
• Spikes that sweep across the interval so no pointwise limit exists
• Spikes that concentrate near a dense set to break naive it is small most places reasoning

Real analysis trains you to ask: what sense of most places do you mean? Almost everywhere, in measure, in L^p, uniformly. Each one has its own stability theorems, and each one has its own counterexamples.

What hypotheses would fix the failure

A counterexample is not complete until you can say precisely what would have prevented it. Here are three standard fixes for limit and integral commute, each addressing a different weakness.

Uniform convergence on a finite interval

If f_n → f uniformly on [0,1], then

lim_{n→∞} ∫_0^1 f_n = ∫_0^1 f.

Proof uses the contract directly:

|∫_0^1 f_n − ∫_0^1 f| ≤ ∫_0^1 |f_n − f| ≤ sup_{x∈[0,1]} |f_n(x) − f(x)|.

Uniform convergence makes that supremum go \to 0. The spike breaks this because the supremum does not go \to 0.

Dominated convergence

Uniform convergence is stronger than necessary. A central result of measure theory is the dominated convergence theorem:

If f_n → f almost everywhere and |f_n| ≤ g for an integrable function g, then ∫ f_n → ∫ f.

The spike breaks this because there is no single integrable function that dominates f̃_n for all n. The heights blow up. Any candidate g would have to exceed n on (0,1/n] for all n, which forces a nonintegrable singularity near 0.

This is the key moral: pointwise convergence does not control tails. Domination is global tail control.

Convergence in L^1

A third fix is to demand that ||f_n − f||_1 → 0. Then ∫ f_n → ∫ f because

|∫ (f_n − f)| ≤ ∫ |f_n − f| = ||f_n − f||_1.

The spike breaks this too: ||f̃_n − 0||_1 = 1 for all n.

So the same example teaches three contracts simultaneously:

• Uniform convergence controls worst-case error
• Domination controls size through an external integrable bound
• L^1 convergence controls average absolute error

Pointwise convergence controls none of these.

A worked diagnostic habit

When you face a claim of the form since f_n → f we can pass the limit through an operation, train yourself to ask:

• Does the operation respond to worst-case values (like sup) or to averages (like integrals)?
• Is my convergence local (pointwise) or global (uniform, L^p, in measure)?
• Do I have a single bound that applies to every n, not just each n separately?
• Where could mass concentrate as n changes?

Spikes answer the last question: mass can concentrate into smaller regions while keeping total area fixed.

Why this is not merely an oddity

The spike sequence is a discrete version of a broader phenomenon: concentration without loss of total quantity. In applications, you see it when approximations become sharply peaked, when gradients become large near interfaces, or when numerical schemes develop boundary layers. Real analysis provides the right language to decide which conclusions are stable under such concentration.

The point is not to memorize this specific sequence. The point is to recognize the mechanism:

• A property verified pointwise can fail to control a global functional
• A global functional can remain stable under concentration
• Stability depends on the right notion of convergence and the right uniform bounds

If you can explain, from the definitions, exactly why f̃_n → 0 pointwise but ∫ f̃_n does not approach 0, you have internalized a large piece of real analysis. Every time you later meet a theorem that allows exchanging limits with integrals, derivatives, or suprema, you will know what obstacle that theorem is removing.