The Real Number System: Completeness and Archimedean Property

⚡ Key Takeaways

The real number system R\mathbb{R} is the unique (up to isomorphism) complete ordered field. Completeness states every nonempty set bounded above has a least upper bound (supremum); the Archimedean property states no real number is infinitely large — for any reals x,y>0x, y > 0 there is a natural number nn with nx>ynx > y. Together they distinguish R\mathbb{R} from Q\mathbb{Q} and underpin all of real analysis. Formalized in the 1870s by Dedekind, Cantor, and Weierstrass.

💡 The Core Insight

Q\mathbb{Q} has holes: the set {x∈Q:x2<2}\{x \in \mathbb{Q} : x^2 < 2\} is bounded above but has no rational least upper bound, since 2∉Q\sqrt 2 \notin \mathbb{Q}. Completeness fills exactly these holes by axiom — it declares that the supremum always exists. The Archimedean property is then the negative statement that we did not accidentally add anything too big: there are no quantities larger than every natural number. The entire edifice of analysis rests on these two facts working together.

Formal Statement

DefinitionOrdered Field

A field (F,+,⋅)(F, +, \cdot) is a set with two operations satisfying the usual axioms (associativity, commutativity, distributivity, identities 0≠10 \ne 1, additive inverses, multiplicative inverses for nonzero elements). An ordered field also carries a total order ≤\le compatible with the operations:

1
a≤b  ⟹  a+c≤b+c,(0≤a ∧ 0≤b)  ⟹  0≤ab.\quad a \le b \implies a + c \le b + c, \qquad (0 \le a \ \wedge\ 0 \le b) \implies 0 \le ab.

Derivation hint: these are the order axioms; everything about inequalities (e.g. squares are nonnegative, a<b∧c>0⇒ac<bca<b \wedge c>0 \Rightarrow ac<bc) follows from them. Here a,b,c∈Fa,b,c \in F. Both Q\mathbb{Q} and R\mathbb{R} are ordered fields.

DefinitionUpper Bound, Supremum

Let S⊆RS \subseteq \mathbb{R} be nonempty. A number MM is an upper bound of SS if s≤Ms \le M for all s∈Ss \in S. The supremum (least upper bound) sup⁡S\sup S is an upper bound LL such that every upper bound MM satisfies L≤ML \le M. Symbolically:

2
L=sup⁡S  ⟺  (∀s∈S, s≤L) ∧ (∀ε>0, ∃s∈S, s>L−ε).\quad L = \sup S \iff \big(\forall s \in S,\ s \le L\big) \ \wedge\ \big(\forall \varepsilon > 0,\ \exists s \in S,\ s > L - \varepsilon\big).

Derivation hint: the first clause says LL is an upper bound; the second says nothing smaller than LL is one — i.e. LL is least, since otherwise L−εL-\varepsilon would bound SS. Dually, inf⁡S\inf S is the infimum (greatest lower bound). Here ε∈R\varepsilon \in \mathbb{R}, ε>0\varepsilon > 0.

TheoremCompleteness Axiom (Least Upper Bound Property)

Every nonempty subset S⊆RS \subseteq \mathbb{R} that is bounded above has a supremum sup⁡S∈R\sup S \in \mathbb{R}.

3
(S≠∅ ∧ ∃M ∀s∈S, s≤M)  ⟹  ∃L∈R, L=sup⁡S.\quad \big(S \ne \varnothing \ \wedge\ \exists M\,\forall s\in S,\ s \le M\big) \implies \exists L \in \mathbb{R},\ L = \sup S.

Derivation hint: this is taken as an axiom for R\mathbb{R}; alternatively it is a theorem once R\mathbb{R} is constructed via Dedekind cuts, where sup⁡S\sup S is literally the union of the cuts.

TheoremArchimedean Property

For every real xx there exists a natural number nn with n>xn > x. Equivalently, for all reals a>0a > 0 and b>0b > 0 there is n∈Nn \in \mathbb{N} with na>bna > b.

4
∀x∈R ∃n∈N, n>x⟺∀ε>0 ∃n∈N, 1n<ε.\quad \forall x \in \mathbb{R}\ \exists n \in \mathbb{N},\ n > x \quad\Longleftrightarrow\quad \forall \varepsilon > 0\ \exists n \in \mathbb{N},\ \tfrac{1}{n} < \varepsilon.

Derivation hint: the two forms are equivalent by taking ε=1/x\varepsilon = 1/x (for x>0x>0); the 1/n→01/n \to 0 form is the one most used in analysis.

TheoremUniqueness of R\mathbb{R}

Any two complete ordered fields are isomorphic as ordered fields. Hence R\mathbb{R} is the complete ordered field, justifying the definite article.

Why It Holds / Derivation

ProofCompleteness implies the Archimedean Property

We show N\mathbb{N} is unbounded in R\mathbb{R}, which is the Archimedean property. 1. Suppose for contradiction N\mathbb{N} is bounded above in R\mathbb{R}. Since N≠∅\mathbb{N} \ne \varnothing, the Completeness Axiom [3] gives L=sup⁡N∈RL = \sup \mathbb{N} \in \mathbb{R}. 2. By the supremum characterization [2] with ε=1\varepsilon = 1: there exists n∈Nn \in \mathbb{N} with

5
n>L−1.\quad n > L - 1.

Derivation hint: L−1<L=sup⁡NL-1 < L = \sup\mathbb N, so L−1L-1 is not an upper bound, hence some element of N\mathbb N exceeds it. 3. Then n+1∈Nn + 1 \in \mathbb{N} and

6
n+1>(L−1)+1=L.\quad n + 1 > (L-1) + 1 = L.

Derivation hint: add 11 to both sides of [5]; closure of N\mathbb N under +1+1 gives a natural number above LL. 4. But n+1∈Nn+1 \in \mathbb{N} and n+1>L=sup⁡Nn+1 > L = \sup \mathbb{N} contradicts LL being an upper bound of N\mathbb{N}. 5. Therefore N\mathbb{N} is unbounded: for any x∈Rx \in \mathbb{R} some n∈Nn \in \mathbb{N} has n>xn > x. ■\blacksquare

∎
ProofQ\mathbb{Q} has a hole — completeness fails in the rationals

Let S={q∈Q:q>0, q2<2}S = \{ q \in \mathbb{Q} : q > 0,\ q^2 < 2\}. 1. SS is nonempty (1∈S1 \in S) and bounded above in Q\mathbb{Q} (e.g. by 22, since q≥2⇒q2≥4>2q\ge 2 \Rightarrow q^2 \ge 4 > 2). 2. Suppose L=sup⁡SL = \sup S existed in Q\mathbb{Q}. Trichotomy gives L2<2L^2 < 2, L2=2L^2 = 2, or L2>2L^2 > 2. 3. L2=2L^2 = 2 is impossible: 2\sqrt 2 is irrational (if L=p/qL = p/q in lowest terms, p2=2q2p^2 = 2q^2 forces both p,qp,q even, contradiction). 4. If L2<2L^2 < 2, choosing h=2−L22L+1h = \frac{2 - L^2}{2L+1} (with 0<h<10<h<1) gives (L+h)2<L2+h(2L+1)=2(L+h)^2 < L^2 + h(2L+1) = 2, so L+h∈SL+h \in S and L+h>LL+h > L — LL is not an upper bound. 5. If L2>2L^2 > 2, a symmetric choice produces a smaller rational upper bound, so LL is not least. 6. All cases fail, so no rational supremum exists. Q\mathbb{Q} violates [3]. ■\blacksquare

∎
🧠 Intuition

Picture the number line. Completeness says: wherever you slice the line into a "left part" and "right part," the cut falls exactly on a number — there is never a microscopic gap where the knife passes through emptiness. In Q\mathbb{Q}, the slice at 2\sqrt 2 falls into a void. The Archimedean property says the number line has no "edge of infinity": start at 00 and march by steps of length aa; you will eventually pass any target bb, no matter how small aa or how large bb. A non-Archimedean line would have points so far out that no finite number of unit steps reaches them, and points so close to 00 that they are positive yet smaller than 1/n1/n for every nn — true infinitesimals.

Figure
The supremum of S={q:q2<2}S=\{q:q^2<2\} fills the gap that Q\mathbb{Q} leaves at 2\sqrt{2}
0 1 2 S = { q : q² < 2 } sup S = √2 no rational sits exactly here — the gap in ℚ left part (q²<2) right part (q²>2)
mathematician · Richard Dedekind | 1831–1916

Dedekind gave the first rigorous construction of R\mathbb{R} in Stetigkeit und irrationale Zahlen (1872) using Dedekind cuts: a real number is identified with a partition of Q\mathbb{Q} into two nonempty sets (A,B)(A, B) where every element of AA is below every element of BB and AA has no greatest element. The cut is the number — 2\sqrt 2 is precisely the cut whose lower set is {q:q<0 or q2<2}\{q : q < 0 \text{ or } q^2 < 2\}. Completeness becomes a theorem: the supremum of a bounded family of cuts is just the union of their lower sets, which is again a cut. Dedekind's insight reduced the mysterious "continuity of the line" to pure set theory over Q\mathbb{Q}, divorcing analysis from geometric intuition once and for all.

Worked Examples

ExampleComputing a supremum from the definition

Given: S={1−1n:n∈N, n≥1}={0, 12, 23, 34,… }S = \left\{ 1 - \dfrac{1}{n} : n \in \mathbb{N},\ n \ge 1 \right\} = \{0,\ \tfrac12,\ \tfrac23,\ \tfrac34,\dots\}. Find sup⁡S\sup S and prove it.

Steps: 1. Guess: terms increase toward 11 but never reach it, so candidate L=1L = 1. 2. Upper bound: 1−1n<11 - \frac1n < 1 for all nn, so 11 bounds SS. ✓ 3. Least: take any ε>0\varepsilon > 0. By the Archimedean property [4] choose nn with 1n<ε\frac1n < \varepsilon. Then

7
1−1n>1−ε,\quad 1 - \tfrac1n > 1 - \varepsilon,

Derivation hint: subtract the inequality 1n<ε\frac1n<\varepsilon from 11, flipping direction. so an element of SS exceeds 1−ε1-\varepsilon. By [2], no number below 11 is an upper bound. 4. Note 1∉S1 \notin S: the supremum need not be attained.

Answer: sup⁡S=1\boxed{\sup S = 1} (and 1∉S1 \notin S, so there is no maximum).

ExampleUsing the Archimedean property numerically

Given: Show there is a natural number nn with 1n<10−6\dfrac{1}{n} < 10^{-6}, and find the smallest such nn.

Steps: 1. The Archimedean property [4] guarantees existence: for ε=10−6>0\varepsilon = 10^{-6}>0 some nn works. 2. Solve 1n<10−6  ⟺  n>106\frac1n < 10^{-6} \iff n > 10^{6}. 3. The smallest integer strictly greater than 106=1,000,00010^6 = 1{,}000{,}000 is

8
n=1,000,001.\quad n = 1{,}000{,}001.

Derivation hint: n=106n=10^6 gives 1/n=10−61/n = 10^{-6}, not strictly less; the strict inequality forces the next integer.

Answer: n=1,000,001\boxed{n = 1{,}000{,}001} (dimensionless count). This concretely exhibits how Archimedean-ness lets 1/n1/n be driven below any positive threshold.

Common Misconceptions

⚠️ Common mistake

"The supremum is the maximum." False in general. sup⁡\sup exists for every bounded-above nonempty set, but the maximum (an element of SS equal to sup⁡S\sup S) need exist. In Example 1, sup⁡S=1∉S\sup S = 1 \notin S. The maximum exists iff sup⁡S∈S\sup S \in S.

⚠️ Common mistake

"Completeness means R\mathbb{R} has no gaps because it's uncountable." Cardinality and completeness are independent. The algebraic numbers are countable yet form a field; the computable reals are countable but not complete. Completeness is an order/topological property, not a counting one. Conversely, one can build non-Archimedean ordered fields of any large cardinality.

CounterexampleCompleteness fails without the least-upper-bound axiom

The field R(t)\mathbb{R}(t) of rational functions in tt, ordered so that tt is larger than every real number (i.e. t>nt > n for all n∈Nn \in \mathbb{N}), is an ordered field that is not Archimedean: N\mathbb{N} is bounded above by tt. Consequently it is not complete. This shows the Archimedean property is not automatic for ordered fields — it is special to R\mathbb{R} and a genuine consequence of completeness, as proved above. Here 1/t1/t is a positive infinitesimal: 0<1/t<1/n0 < 1/t < 1/n for every nn.

connection

Completeness is the hinge on which all of analysis turns. It is logically equivalent to: the Monotone Convergence Theorem (bounded monotone sequences converge), the Bolzano–Weierstrass Theorem (bounded sequences have convergent subsequences), the Nested Interval Property, Cauchy completeness plus the Archimedean property, and the Heine–Borel Theorem. Each can serve as the starting axiom; the others become theorems. Completeness also makes R\mathbb{R} connected as a topological space and guarantees the Intermediate Value Theorem and Extreme Value Theorem. The Archimedean property links R\mathbb{R} to the density of Q\mathbb{Q} in R\mathbb{R}.

Real-World Applications

🔬 Where this appears

Historical Development

⏳ History
- c. 450 BCE — The Pythagoreans discover incommensurability (2\sqrt2 irrational), the first crack revealing that ratios of integers do not fill the line.
- c. 350 BCE — Eudoxus' theory of proportion (Euclid, Elements Book V) handles incommensurable magnitudes; his definition of equal ratios anticipates Dedekind cuts by two millennia.
- 287–212 BCE — Archimedes states the axiom now bearing his name (attributing it to Eudoxus) and uses it in his method of exhaustion.
- 1817 — Bolzano gives an early (largely unread) proof of the intermediate value theorem, implicitly invoking completeness.
- 1821 — Cauchy's Cours d'analyse introduces rigorous limits and "Cauchy sequences," but without a construction of R\mathbb{R} the completeness gap remains.
- 1872 — Dedekind (cuts) and Cantor & Heine (Cauchy sequences of rationals) independently construct R\mathbb{R} from Q\mathbb{Q}, giving completeness a firm foundation.
- 1872 — Weierstrass and his school formalize the ε\varepsilon–δ\delta method, relying on the least-upper-bound property.
- 1899 — Hilbert axiomatizes R\mathbb{R} as a complete Archimedean ordered field (his "axiom of completeness").
- 1960 — Abraham Robinson founds non-standard analysis, constructing the non-Archimedean hyperreals and legitimizing infinitesimals.

Subtleties & Limits

open-problem

Completeness and Archimedean-ness are first-order-elusive: the least-upper-bound property is a second-order statement (it quantifies over all subsets), and by the Löwenheim–Skolem theorem no first-order theory pins R\mathbb{R} down uniquely — there exist non-isomorphic models of the first-order theory of real-closed fields, some non-Archimedean. This is the root of non-standard analysis. Open and ongoing themes include: (i) the precise logical strength of completeness in reverse mathematics, where many analysis theorems sit at the level ACA0\mathsf{ACA}_0 or WKL0\mathsf{WKL}_0; (ii) constructive real analysis, where the classical least-upper-bound principle fails and must be replaced by weaker located-supremum notions (Bishop's school); and (iii) whether physical spacetime is genuinely a real continuum or only approximated by one — an empirically open question at the Planck scale.

See Also

Dedekind Cuts · Bolzano–Weierstrass Theorem · Nested Interval Property · Density of Rationals · Non-Standard Analysis · Supremum and Infimum · Monotone Convergence Theorem · Real-Closed Fields