Epsilon-Delta Definition — Mathematics Knowledge System

Overview

The epsilon-delta (ε-δ) definition is the rigorous formalization of what it means for a function to approach a limit. Instead of relying on the intuitive but vague idea of a variable 'getting arbitrarily close' to a value, it converts the statement into a precise back-and-forth challenge: for every tolerance ε (epsilon) someone names around the limit L, you must be able to produce a corresponding neighborhood δ (delta) around the input a such that every x within δ of a (excluding a itself) makes f(x) land within ε of L. This definition, due to Bolzano, Cauchy, and finally Weierstrass, replaced two centuries of informal 'infinitesimal' reasoning and put the whole of calculus and real analysis on solid logical footing.

Intuition

Imagine an adversarial game. Someone (the skeptic) challenges you: 'Prove f(x) approaches L as x approaches a. I bet you can't keep f(x) within 0.001 of L.' You respond by finding a small window around a — say, within 0.0003 — such that any x in that window (other than a itself) forces f(x) within their requested 0.001 of L. If you can always win this game no matter how small a tolerance (ε) the skeptic names, by producing a suitable window (δ), then the limit truly is L. The crucial subtlety is the order of the challenge: the skeptic picks ε FIRST, and only then do you get to pick δ in response — your δ is allowed to depend on their ε (typically δ shrinks as ε shrinks). You never get to see 'a delta good for all epsilons at once' handed to you first.

Interactive Graph

Formal Definition

Definition

Let f be a function defined on an open interval containing a, except possibly at a itself. We say the limit of f(x) as x approaches a is L, written lim(x→a) f(x) = L, if:

\forall \varepsilon > 0,\ \exists \delta > 0 \text{ such that } \forall x,\ \big(0 < |x-a| < \delta \implies |f(x)-L| < \varepsilon\big)

For every tolerance ε around L there exists a corresponding neighborhood δ around a (excluding a) inside which f is guaranteed to be within ε of L

ε-δ definition of a limit

0 < |x - a| < \delta

x is within δ of a, but not equal to a — the punctured neighborhood; the limit does not care about (or require) the value of f at a itself

|f(x) - L| < \varepsilon

f(x) lands strictly within the ε-tolerance band around the claimed limit L

Notation

Notation	Meaning
$\varepsilon$	An arbitrary positive tolerance around the limit value L, chosen first (universally quantified)
$\delta$	A positive tolerance around the point a, chosen in response to ε (existentially quantified, may depend on ε)
$\forall$	'For all' — universal quantifier; ε ranges over all positive reals
$\exists$	'There exists' — existential quantifier; asserts at least one valid δ can be found
$0 < \|x-a\| < \delta$	The punctured δ-neighborhood of a: all x within δ of a except a itself
$\delta(\varepsilon)$	Notation emphasizing that the choice of δ is a function of the given ε

Derivation

Proving, from the ε-δ definition alone, that lim(x→2) (3x + 1) = 7. The strategy in any such proof is to work backward from |f(x) - L| < ε to discover how large δ must be, then present the argument forward.

|f(x) - L| = |(3x+1) - 7| = |3x - 6| = 3|x-2|

Simplify the target quantity in terms of |x - a|

3|x-2| < \varepsilon \iff |x-2| < \frac{\varepsilon}{3}

Solve the inequality for |x-2| to see what δ needs to be

\text{Choose } \delta = \frac{\varepsilon}{3}

This is the 'scratch work' — the formal proof restates this choice up front

0 < |x-2| < \delta = \tfrac{\varepsilon}{3} \implies |3x-6| = 3|x-2| < 3\cdot\tfrac{\varepsilon}{3} = \varepsilon

Forward verification: assuming the δ-bound, the ε-bound follows exactly as required

Proofs

lim(x→2) (3x + 1) = 7

$\text{Let } \varepsilon > 0 \text{ be given.}$ (The proof must work for an arbitrary, unspecified positive ε — this is the universal quantifier)
$\text{Choose } \delta = \varepsilon/3.$ (Determined by scratch work: solving 3|x-2|<\varepsilon for |x-2|)
$\text{Suppose } 0 < |x-2| < \delta.$ (Assume the hypothesis of the implication we must prove)
$|f(x)-7| = |(3x+1)-7| = |3x-6| = 3|x-2|$ (Algebraic simplification of the quantity to be bounded)
$3|x-2| < 3\delta = 3\cdot\frac{\varepsilon}{3} = \varepsilon$ (Multiply the assumed bound |x-2|<\delta by 3, then substitute δ=ε/3)
$\therefore |f(x)-7| < \varepsilon \text{ whenever } 0<|x-2|<\delta.$ (This is exactly the ε-δ condition, so the limit is confirmed to equal 7)

Uniqueness of limits

$\text{Suppose } \lim_{x\to a}f(x) = L_1 \text{ and } \lim_{x\to a}f(x)=L_2 \text{ with } L_1 \neq L_2.$ (Assume, for contradiction, that a limit could have two distinct values)
$\text{Let } \varepsilon = \frac{|L_1 - L_2|}{2} > 0.$ (Since L₁ ≠ L₂, this ε is a well-defined positive number — half the gap between the two claimed limits)
$\exists\, \delta_1>0: 0<|x-a|<\delta_1 \implies |f(x)-L_1|<\varepsilon, \text{ and similarly } \delta_2 \text{ for } L_2.$ (Apply the ε-δ definition to each of the two assumed limits with this specific ε)
$\text{Let } \delta = \min(\delta_1,\delta_2). \text{ Pick any } x \text{ with } 0<|x-a|<\delta.$ (Such x satisfies both δ-conditions simultaneously)
$|L_1 - L_2| \le |L_1-f(x)| + |f(x)-L_2| < \varepsilon+\varepsilon = |L_1-L_2|$ (Triangle inequality, then substituting the two ε-bounds; this yields |L₁-L₂| < |L₁-L₂|)
$\text{Contradiction.} \therefore L_1 = L_2.$ (A quantity cannot be strictly less than itself, so the assumption L₁≠L₂ is false)

Properties

Symmetry of roles

\varepsilon \text{ is chosen first (universally); } \delta \text{ is chosen second (existentially) and may depend on } \varepsilon.

Condition: Reversing the quantifier order (∃δ ∀ε) produces a fundamentally different, much stronger and usually false statement.

Irrelevance of f(a)

0 < |x-a| \text{ excludes } x=a, \text{ so the limit does not depend on (or require) } f(a) \text{ to be defined.}

Non-uniqueness of δ

\text{If } \delta \text{ works for a given } \varepsilon, \text{ then any smaller } \delta' \in (0,\delta) \text{ also works.}

Example: If δ=ε/3 works, δ=ε/10 works too — proofs conventionally report the largest convenient choice.

Equivalence to sequential limits

\lim_{x\to a} f(x) = L \iff \text{for every sequence } x_n \to a \text{ with } x_n \neq a,\ f(x_n) \to L.

Condition: This is the Heine (sequential) characterization of the limit, a common alternative proof technique.

Applications

Numerical analysis and floating-point error bounds are literally ε-δ style guarantees: an algorithm's output is proven to stay within a tolerance ε for sufficiently fine step size or precision δ.

Worked Examples

From the general derivation, δ = ε/3 always works for this function.
$\delta = \frac{0.3}{3} = 0.1$
Check: if 0<|x-2|<0.1, then |3x+1-7| = 3|x-2| < 3(0.1) = 0.3 = ε.
$3(0.1) = 0.3$

Answer: δ = 0.1 works for ε = 0.3

Practice Problems

Difficulty 5/10

Find a δ (in terms of ε) that proves lim(x→1) (2x - 3) = -1.

Difficulty 4/10

In the ε-δ definition, which of the following correctly describes the quantifier order?

There exists a single δ that works for all ε simultaneouslyFor every ε, there exists a δ (which may depend on ε) that worksFor every δ, there exists an ε that worksε and δ are chosen independently and in either order

Difficulty 7/10

Prove using ε-δ that lim(x→4) √x = 2.

Common Mistakes

Common Mistake

Swapping the quantifier order and treating 'there exists δ such that for all ε...' as equivalent to the actual definition.

The correct order is ∀ε ∃δ — δ is chosen AFTER and IN RESPONSE TO ε. Reversing it (∃δ ∀ε) would require one single δ to work for every tolerance no matter how small, which is a far stronger (and for non-constant functions, false) claim.

Common Mistake

Believing the definition requires or uses the value f(a).

The condition 0 < |x-a| strictly excludes x = a. The limit describes behavior arbitrarily close to a but never at a, so f(a) can be undefined, or even different from L, without affecting the limit.

Common Mistake

Thinking any δ that happens to work is 'the' answer, or that δ must be found without any algebraic scratch work.

δ is never unique — any smaller positive value also works. The scratch work (solving |f(x)-L|<ε backward for |x-a|) is exactly how δ is discovered; it is legitimate and expected, not a shortcut to avoid.

Quiz

Which statement is the correct ε-δ definition of lim(x→a) f(x) = L?

∃δ>0 ∀ε>0: 0<|x-a|<δ ⟹ |f(x)-L|<ε∀ε>0 ∃δ>0: 0<|x-a|<δ ⟹ |f(x)-L|<ε∀ε>0 ∀δ>0: 0<|x-a|<δ ⟹ |f(x)-L|<ε∀ε>0 ∃δ>0: |f(x)-L|<ε ⟹ 0<|x-a|<δ

For f(x) = 3x+1 near a=2, if δ=ε/3 is shown to satisfy the definition, which of the following also necessarily satisfies it for the same ε?

δ = ε/2 (larger)δ = ε/6 (smaller)Any δ > ε/3None — δ=ε/3 is the unique valid choice

Flashcards

1 / 4

Historical Background

Newton and Leibniz's calculus worked but leaned on 'infinitesimals' — quantities treated as nonzero yet smaller than any positive real number, a notion George Berkeley mocked in 1734 as 'ghosts of departed quantities.' Bernard Bolzano gave an early rigorous approach to limits in the 1810s, largely unnoticed at the time. Augustin-Louis Cauchy's 1821 Cours d'Analyse described limits in terms of quantities becoming 'as small as one wishes,' a phrase still verbal rather than fully symbolic. It was Karl Weierstrass, lecturing at the University of Berlin in the 1850s–1861, who converted this into the fully symbolic ∀ε ∃δ statement used today, banishing infinitesimals from rigorous analysis (though they would resurface a century later, made logically sound, in Abraham Robinson's nonstandard analysis).

1817
Bolzano gives an early rigorous definition of continuity and limit, published with little immediate impact
Bernard Bolzano
1821
Cauchy's Cours d'Analyse describes limits verbally in terms of quantities becoming arbitrarily small
Augustin-Louis Cauchy
1861
Weierstrass formulates the modern symbolic ε-δ definition in his Berlin lectures
Karl Weierstrass
1960s
Abraham Robinson develops nonstandard analysis, giving infinitesimals a rigorous logical foundation as an alternative to ε-δ
Abraham Robinson

Summary

The ε-δ definition formalizes lim(x→a) f(x)=L as: ∀ε>0 ∃δ>0 such that 0<|x-a|<δ ⟹ |f(x)-L|<ε.
Quantifier order is essential: ε is chosen first (for all), δ is chosen second (there exists) and may depend on ε — reversing the order changes the meaning.
The condition 0<|x-a| excludes x=a, so the limit never depends on f(a) itself.
δ is never unique: any smaller positive value that satisfies the implication also works.
Standard proofs work backward from the ε-inequality to discover δ (scratch work), then present the argument forward from the assumed δ-bound to the required ε-bound.

References

WebsiteWikipedia — Limit of a function
BookRudin, W. Principles of Mathematical Analysis, 3rd ed. Ch. 4.
BookWeierstrass, K. (1861 lectures, Berlin) — the first fully symbolic ε-δ formulation, as reconstructed in later editions of his students' notes.