Uni Reading List
Study guides /

Real analysis: make the quantifiers do the work

A convincing graph and a correct limit at every fixed point can still miss the statement you need. Choose a text that fits your proof experience, then track which choices may depend on the point and which must work across the whole domain.

Course topic: Real Analysis

Study route and exercise: University Reading List editorial project. How we curate · Sources checked:

Download blank worksheet (Markdown) ↓
On this page

Scope and responsibility

A first proof-based route through sequences, function limits and uniform convergence on a real interval. Lebl means the sixth major edition, volume I; Abbott the 2015 second edition; Bartle–Sherbert the 2011 fourth edition. This is single-variable analysis preparation, with no assumption of measure theory. Later metric-space or multivariable material is outside this six-output route.

Choose your starting point

I need a free proof course

Choose Lebl after a basic introduction to proofs; arrange feedback because the author intentionally supplies no exercise solutions manual.

I need motivation before formal detail

Choose Abbott when discussion examples help you understand what a definition is protecting against.

I want a systematic traditional spine

Choose Bartle–Sherbert for a conventional progression from preliminaries and real numbers to limits and continuity.

Compare the three books

Basic Analysis: Introduction to Real Analysis

Jiří Lebl · Edition: 6

Free first analysis course

Choose it when
You can write quantified statements and want a complete volume without a purchase.
Use another route when
You need a complete public answer key or want multivariable calculus before real-line foundations.
Selected reading
Volume I: Sequences and Series, Continuous Functions, then Sequences of Functions; use Real Numbers for completeness.
Limits and trade-offs
No solutions manual: checking each implication and obtaining feedback are part of the work. Volume II is a continuation.

Try before choosing

Using a sequence-limit definition, prove 1/(n+1)→0 with an explicit index choice.

Check your work: Give N depending on epsilon, then prove the inequality for every n≥N; a list of decimals is insufficient.

Official free reading ↗

Book record and access options →

Understanding Analysis

Stephen Abbott · Edition: 2 · 2015

Motivated proof development

Choose it when
You need examples that make approximation and finite-to-infinite pitfalls concrete.
Use another route when
You want only computational calculus or expect instructor solutions to be freely available.
Selected reading
Sequences and Series, Functional Limits and Continuity, and Sequences and Series of Functions.
Limits and trade-offs
Discussion and projects reward slower investigation; they are not a shortcut around writing the proofs.

Try before choosing

For x^n on [0,1], write the pointwise limit and locate the failure of a uniform claim.

Check your work: Treat x=1 separately and exhibit points near 1; do not infer uniformity from a graph at fixed resolution.

Book record and access options →

Introduction to Real Analysis

Robert G. Bartle; Donald R. Sherbert · Edition: 4 · 2011

Systematic first-course reference

Choose it when
You prefer definitions and a regular theorem/exercise sequence across the real line.
Use another route when
You are looking for a graduate measure-theory or general multivariable-analysis text.
Selected reading
Preliminaries, The Real Numbers, Sequences and Series, Limits, Continuous Functions; later sequences of functions as required.
Limits and trade-offs
The conventional structure is easy to follow by topic, but you must supply examples that distinguish similar hypotheses.

Try before choosing

For f_n below, produce an upper bound independent of x and explain the role of a supremum.

Check your work: The same N must control every x in the domain. A separate N(x) proves a weaker statement.

Book record and access options →

Check your preparation

Can I negate a quantified limit statement?

Write the order of epsilon, N, n and x. Uniform failure means one positive tolerance defeats every proposed common N.

Can I manipulate inequalities before choosing N?

For 1/(n+1)<epsilon, choose N=ceil(1/epsilon). For n≥N the inequality is strict; keep epsilon positive.

Can I use calculus without mistaking it for a proof of convergence?

A derivative can locate a global maximum after endpoints and derivative signs are checked. Uniform convergence still needs a bound tending to zero.

A six-step study plan

These are your own study marks, not an assessment. They are not sent to this site.

  1. Write two definitions

    Selected reading: Your chosen sequence/function-limit units.

    What to do: Produce pointwise and uniform convergence definitions side by side, marking permitted dependencies.

    Check your work: Only the pointwise definition permits N to depend on x.

  2. Prove a scalar limit

    Selected reading: Lebl Sequences and Series; Abbott or Bartle matching unit.

    What to do: Write a complete epsilon proof for 1/(n+1).

    Check your work: The final implication applies to every n≥N, not only n=N.

  3. Find the global error

    Selected reading: Your chosen derivative and sequences-of-functions units.

    What to do: Differentiate x^n(1−x), check endpoints and produce its exact maximum on [0,1].

    Check your work: An interior critical point alone does not establish the maximum.

  4. Convert a bound into a tolerance

    Selected reading: Sequences of functions.

    What to do: Use a bound independent of x to provide a valid common N for the worked example.

    Check your work: The proof explicitly includes x=0 and x=1.

  5. Stress-test the hypothesis

    Selected reading: The same uniform-convergence unit.

    What to do: Solve the n-scaled follow-up; give both a fixed-point argument and a moving-point obstruction.

    Check your work: A nonzero lower bound along x_n defeats uniformity even though each fixed point converges.

  6. Review a false proof

    Selected reading: Definitions and your own two solutions.

    What to do: Annotate the claim “each x has a suitable N, so take one N for all x.”

    Check your work: Identify the unjustified quantifier interchange and replace it by a supremum estimate or counterexample.

Worked exercise

A moving maximum can still shrink uniformly

For integers n≥1, let f_n(x)=x^n(1−x) on [0,1]. Find the pointwise limit and the exact maximum error. Prove uniform convergence to zero and give a common index for a tolerance epsilon>0.

Assumptions

Real-valued functions on the whole closed interval [0,1]; ordinary pointwise and supremum-norm convergence. Each n is a positive integer.

  1. For fixed 0≤x<1, x^n→0 and 0≤1−x≤1; at x=1 the function is identically zero. The pointwise limit is therefore zero everywhere.
  2. For 0<x<1, f_n′(x)=x^(n−1)[n−(n+1)x]. It is positive before x=n/(n+1) and negative after it; both endpoints give zero. The global maximum is n^n/(n+1)^(n+1).
  3. Rewrite that maximum as (n/(n+1))^n/(n+1)≤1/(n+1). Choose N=ceil(1/epsilon); for every n≥N and every x, |f_n(x)|≤1/(n+1)<epsilon.

Answer and interpretation

The pointwise and uniform limits are both zero. The maximum is attained at x=n/(n+1), with value n^n/(n+1)^(n+1). N=ceil(1/epsilon) is a valid common index.

Test the result

For n=1 the maximum is 1/4 at x=1/2. For n=2 it is 4/27 at x=2/3. The maximizing point approaches 1, but its height tends to zero; motion of a peak alone does not disprove uniform convergence.

Change an assumption

A growing prefactor changes the quantifier problem

Keep the domain [0,1] but replace f_n by g_n(x)=n x^n(1−x). Determine pointwise and uniform convergence to zero. Test x_n=1−1/(2n), and justify the fixed-point limit rather than assuming it.

  1. At x=0 and x=1 the value is zero. For fixed 0<x<1, set a_n=n x^n. Its successive ratio is x(1+1/n). Choose q with x<q<1; eventually the ratio is at most q, so a_n is bounded by a decaying geometric sequence. Thus g_n(x)→0 at each fixed point.
  2. At x_n=1−1/(2n), g_n(x_n)=(1/2)(1−1/(2n))^n. Bernoulli’s inequality gives (1−1/(2n))^n≥1−n/(2n)=1/2, so this value is at least 1/4 for every n.
  3. Take epsilon=1/8. For every proposed N, choose n≥max(N,1) and the corresponding x_n. Its error exceeds epsilon, so no common index controls the entire interval.

Answer and interpretation

g_n converges pointwise to zero but not uniformly on [0,1]. The moving points retain an error of at least 1/4.

Test the result

For n=1, x_1=1/2 and g_1(x_1)=1/4. The obstruction is a sequence of changing points; it does not contradict convergence at any fixed x. A finite numerical grid may miss a narrow peak.

Common mistakes

An index with hidden dependence

Writing N(x,epsilon) and later calling it a common N changes the claim. Keep dependence visible.

Only checking stationary points

Compare endpoints and derivative signs before reporting a global maximum.

Using a finite plot as a proof

A fixed grid cannot rule out peaks moving between samples as n changes. Use a bound or an explicit witness.

Before moving on

  • Write a quantified epsilon proof with a common index.
  • Explain why the scaled family defeats uniformity without defeating pointwise convergence.
  • Choose a text and a proof-feedback plan that match your readiness.

Sources and edition checks

Sources checked:

These are proposed study activities, not official course assignments or a claim that every book has been read cover to cover.

How we curate →