University of Toronto · FACULTY OF MATHEMATICS

MAT148H1 Chap.4 The Epsilon-Delta Definition of a Limit

- one subject, every graph, every model, every mark
7 Chapters4-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 4 of 9 · MAT148H1

The Epsilon-Delta Definition of a Limit

Reading the definition as a challenge

The statement lim f(x) = L as x → a means: for every ε > 0 there exists δ > 0 such that for all x, if 0 < |x − a| < δ then |f(x) − L| < ε. The absolute values are distances, so |x − a| < δ says x sits within δ of a, and |f(x) − L| < ε says the output sits within ε of L.

Think of a challenger who names any tolerance ε, however tiny, and your job is to answer with a window δ that guarantees every input in the window, except a itself, produces an output inside the tolerance. Because δ is chosen after ε, a correct answer is a formula in ε rather than a single number.

The function only has to be defined on an open interval around a with a possibly removed, so the value f(a), or the lack of one, never affects the limit.

From rough work to a written proof

Every ε-δ proof has two layers.

The rough work starts from |f(x) − L|, factors out |x − a|, and bounds the remaining factor by a constant k after a preliminary restriction such as |x − a| < 1. For example, near 2 the expression x² + x − 6 factors as (x − 2)(x + 3); the restriction |x − 2| < 1 places x between 1 and 3, so |x + 3| < 6. Choosing δ = min(1, ε/6) then makes both restrictions hold simultaneously.

The written proof reverses the order: fix ε, state δ explicitly, take an arbitrary x in the punctured window, and chain inequalities until the error is below ε. A good proof is self-contained, defines every variable before use, and never assumes its conclusion.

For rational functions an extra step bounds the denominator from below so that its reciprocal stays bounded above.

Showing a limit fails

Negating the definition swaps each quantifier: lim f(x) ≠ L exactly when there exists ε > 0 such that for every δ > 0 there is some x with 0 < |x − a| < δ and |f(x) − L| ≥ ε. In plain terms, one fixed tolerance defeats every window.

The usual disproof names a specific ε, lets δ be arbitrary, and builds a witness point from δ, often something like a + min(δ, 1)/2 so that it lies strictly inside the window.

To show that no limit exists at all, a universal quantifier over L goes in front, and ε may then depend on L.

Why the definition matters for the rest of the course

Every limit law, the uniqueness of limits, continuity, and the derivative itself rest on this sentence.

Uniqueness, for instance, follows by taking ε as half the distance between two candidate limits: the two ε bands are disjoint, and the graph near a cannot lie in both. Equivalent reformulations, such as replacing every real M by every integer M in the definition of an infinite limit, are proved the same way, by fixing the arbitrary object first and choosing the window second.

Continuity at a uses the identical sentence with L replaced by f(a), so this proof shape carries straight into the continuity arguments that follow. A practical habit is to test any proposed δ numerically: with ε = 0.6 the rule above gives δ = 0.1, and x = 2.09 produces an error near 0.458, comfortably inside the tolerance.

In this chapter

What this chapter covers

  • 01

    Quantifier order in the ε-δ definition

  • 02

    Epsilon bands and delta windows on a graph

  • 03

    Factor and bound routine for polynomials

  • 04

    Choosing δ = min(1, ε/k)

  • 05

    Negation of the limit definition

  • 06

    Witness points for limits that fail

  • 07

    Uniqueness of limits

Worked example · free

An ε-δ proof that x² + x tends to 6 at 2

Q [6 marks]. AskSia-authored practice. Prove directly from the ε-δ definition that the limit of x² + x as x → 2 equals 6. (6 marks: AskSia study weighting, not a University marking scheme.)
The horizontal strip has half-height ε around L; the vertical strip has half-width δ around a.The horizontal strip has half-height ε around L; the vertical strip has half-width δ around a. The curve leaves the band at different distances on each side, so δ is the smaller one. The hollow dot marks that f(a) itself is never tested. L+ε L L−ε a−δ a a+δ window inside, graph inside band
The horizontal strip has half-height ε around L; the vertical strip has half-width δ around a. The curve leaves the band at different distances on each side, so δ is the smaller one. The hollow dot marks that f(a) itself is never tested.
  • 2Rough work: |x² + x − 6| = |x − 2| |x + 3|. If |x − 2| < 1 then 1 < x < 3, so 4 < x + 3 < 6 and |x + 3| < 6.
  • 1Let ε > 0 and choose δ = min(1, ε/6), so δ ≤ 1 and δ ≤ ε/6.
  • 1Let x satisfy 0 < |x − 2| < δ. Since δ ≤ 1, we have 1 < x < 3 and therefore |x + 3| < 6.
  • 2Then |x² + x − 6| = |x − 2| |x + 3| < (ε/6)(6) = ε. Since ε was arbitrary, the limit equals 6.
With δ = min(1, ε/6), every x satisfying 0 < |x − 2| < δ gives |x² + x − 6| < ε, so the limit is 6.
Sia tip — Box the line that states δ as an exact formula in ε before you introduce x.
Glossary

Key terms

Epsilon Band
An epsilon band is the horizontal strip of points whose height lies strictly between L − ε and L + ε. A limit statement requires the graph near a to stay inside this strip for every positive ε.
Delta Window
A delta window is the set of inputs x with 0 less than |x − a| less than δ, an open interval around a with the centre removed. The prover chooses its radius after the tolerance is given.
Punctured Neighbourhood
A punctured neighbourhood is an open interval around a point with that point deleted. Limits only inspect inputs in such a set, so the value of the function at the point itself is irrelevant.
Witness Point
A witness point is a specific input, built from an arbitrary δ, lying inside the delta window yet sending the output outside the epsilon band. Producing one for every δ proves that a proposed limit fails.
Minimum Rule
The minimum rule is the choice δ = min(δ₁, δ₂), which ensures that two separate restrictions on |x − a| hold simultaneously, typically one bounding a factor and one shrinking the error.
FAQ

The Epsilon-Delta Definition of a Limit FAQ

How do you find delta in an epsilon delta proof?

Start from |f(x) − L| and factor out |x − a|. Restrict |x − a| less than 1 to bound the leftover factor by a constant k, then take δ = min(1, ε/k). That choice makes the bound and the smallness work together in the final inequality.

Why is the condition 0 less than |x − a| included in the limit definition?

It deletes the point a from the set of tested inputs, so the limit describes behaviour near a rather than at a. Functions with holes, or with a value at a that differs from nearby values, still have well-defined limits under this convention.

How do you prove that a limit does not equal a given number?

Use the negation: pick one positive tolerance, let the window radius be arbitrary, and construct an input inside that window whose output misses the target by at least the tolerance. The input may depend on the radius, but the tolerance must be fixed first.

Can delta be any number smaller than the one you found?

Yes. Shrinking the window only removes inputs from consideration, so every inequality that held for the larger radius still holds for the smaller one. Proofs report one convenient choice, and different correct proofs often state different but equally valid formulas.

Study strategy

Exam move

Each day, write the definition from memory with the quantifiers in order, then complete one forward proof and one negation proof. For forward proofs use a fixed checklist: factor, cap by 1, compute k, state δ = min(1, ε/k), write up. Test your δ numerically at one ε and one x near a. For negations, always state ε first and construct the witness explicitly from δ.

Working through The Epsilon-Delta Definition of a Limit in MAT148H1? Sia is AskSia’s AI Mathematics tutor — ask any MAT148H1 The Epsilon-Delta Definition of a Limit question and get a clear, step-by-step explanation grounded in how MAT148H1 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 21 of your University of Toronto subjects - and 1,000+ Bibles across every Australian university.
Sia - your MAT148 tutor, unlimited, worked the way the exam marks it
The full 4-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works