The Archimedean Property

Can a positive number be so small, or a number so large, that it escapes the reach of the natural numbers entirely? Intuition says no. However tiny a positive $\varepsilon$ is, adding it to itself enough times should eventually overtake any fixed real number. Yet intuition is not proof. The history of mathematics contains number systems — the hyperreals, the surreals, and other ordered fields — where this "no infinitesimal quantities" intuition genuinely fails. The Archimedean Property is the precise statement that the real numbers, as constructed via the Completeness Axiom, do not admit such rogue elements. The natural numbers are unbounded above in $\mathbb{R}$ . This property matters because it is the quiet engine behind an enormous number of estimates in analysis. Every time you say "choose $n$ large enough that $1/n < \varepsilon$ ," or "choose $n$ so that $2^{-n}$ is smaller than any prescribed tolerance," you are invoking the Archimedean Property. It is what rules out the existence of infinitesimals in $\mathbb{R}$ . It is the single fact that converts the Completeness Axiom — an abstract statement about suprema — into concrete, usable numerical control. The big picture: completeness of $\mathbb{R}$ is what makes $\mathbb{R}$ different from $\mathbb{Q}$ . But completeness alone does not obviously tell you that $\mathbb{N}$ is unbounded. That requires an argument. The argument is a clean, short application of the Epsilon Characterization of Suprema. Once established, the Archimedean Property unlocks a cascade of equivalent formulations. These concern $1/n$ , about $\lfloor x \rfloor$ , about spacing of reals. It also feeds directly into the proof that $\mathbb{Q}$ is dense in $\mathbb{R}$ . It feeds into constructions of square roots and other limiting objects. This article gives the precise statement, and a complete proof from completeness. It offers several logically equivalent restatements, each proved equivalent. It includes worked numerical examples, and a counterexample showing what fails in a non-Archimedean ordered field. Finally, it surveys where the property is used throughout analysis, geometry, and beyond.

Key takeaways

The Archimedean Property says: for every real $x$ , there exists a natural number $n$ with $n > x$ .
Equivalently: for every $\varepsilon > 0$ , there is $n \in \mathbb{N}$ with $1/n < \varepsilon$ .
It is not an axiom of $\mathbb{R}$ separately — it is a theorem, derived from the Completeness Axiom.
The proof is a proof by contradiction using the Epsilon Characterization of Suprema applied to $\mathbb{N}$ .
Ordered fields can fail to be Archimedean (e.g. fields with infinitesimals); completeness is what rescues $\mathbb{R}$ .
Key idea: completeness forbids $\mathbb{N}$ from having a supremum, because $\sup \mathbb{N} - 1$ could never be a valid upper-bound threshold.

Where this lives today

The Archimedean Property is not a museum piece — it is used, often invisibly, every time an analyst manipulates an $\varepsilon$ – $N$ or $\varepsilon$ – $\delta$ argument.

Real analysis: the standard proof that $\lim_{n\to\infty} 1/n = 0$ is a direct restatement of form (2); virtually every convergence proof in a first analysis course that "picks $N$ large enough" implicitly invokes it.
Density of $\mathbb{Q}$ in $\mathbb{R}$ : the proof that between any two reals lies a rational uses the Archimedean Property to find a denominator $n$ large enough that $1/n$ is smaller than the gap between the two numbers, then uses the floor function (form (4)) to locate the numerator.
Construction of square roots and general $n$ th roots: the standard supremum-based construction of $\sqrt{2} = \sup\{x \in \mathbb{Q} : x^2 < 2\}$ relies on being able to shrink error terms $1/n$ arbitrarily, invoking the Archimedean Property to complete the "no gap" argument that the supremum squares to exactly $2$ .
Measure theory and integration: the definition of Lebesgue measure via countable covers by intervals of total length within $\varepsilon$ of the infimum relies on Archimedean control of tolerances.
Numerical analysis and floating-point arithmetic: algorithms with stopping criteria like "iterate until the error is below $10^{-8}$ " are guaranteed to terminate, in the idealized real-number model, precisely because $1/n \to 0$ .
Non-Archimedean mathematics as a contrast case: in $p$ -adic analysis and nonstandard analysis, one deliberately studies fields (the $p$ -adic numbers, the hyperreals) where the Archimedean Property fails or is replaced by a different valuation structure — these fields are central to modern number theory (via $p$ -adic numbers) and to rigorous treatments of infinitesimal calculus (via nonstandard analysis), and understanding what "Archimedean" buys you in $\mathbb{R}$ clarifies what is lost or gained in these alternative settings.
Computer science / cryptography: models of computation using ordered fields for verified numerics, and complexity-theoretic arguments bounding "the number of steps until a quantity crosses a threshold," rest on the same finite-step guarantee that the Archimedean Property provides.

Ask the tutor: How does the Archimedean Property get used, step by step, in the standard proof that the rationals are dense in the reals?

F

Intuition

Picture the real line, and imagine trying to "fence in" all the natural numbers with some finite upper bound $M$ . The Archimedean Property says this is impossible: no matter how large $M$ is, you can always find a natural number that steps over it. This feels almost too obvious to need proof — of course $1, 2, 3, \ldots$ marches off to infinity — but the subtlety is that obviousness is not a substitute for a derivation from the axioms. In an arbitrary ordered field, there is no a priori reason $\mathbb{N}$ cannot be bounded; the reason it cannot happen in $\mathbb{R}$ specifically is that $\mathbb{R}$ is complete, and completeness is precisely the axiom that rules out "gaps" where a bounded set could fail to have a genuine least upper bound. The mechanism of the proof is elegant: assume for contradiction that $\mathbb{N}$ is bounded above. Then, by completeness, $\mathbb{N}$ has a least upper bound $\alpha = \sup \mathbb{N}$ . But $\alpha - 1$ is smaller than $\alpha$ , so it is not an upper bound for $\mathbb{N}$ (by the very definition of "least" upper bound, or via the Epsilon Characterization of Suprema with $\varepsilon = 1$ ). Hence some natural number $n$ satisfies $n > \alpha - 1$ , which forces $n+1 > \alpha$ — but $n + 1 \in \mathbb{N}$ , contradicting that $\alpha$ is an upper bound for $\mathbb{N}$ at all. The natural numbers, in effect, "outrun" any candidate ceiling by exploiting the freedom to add $1$ . This same mechanism is what lets you push $1/n$ below any positive tolerance $\varepsilon$ : as $n$ ranges over $\mathbb{N}$ without bound, $1/n$ is squeezed toward $0$ without bound, and no fixed positive $\varepsilon$ can serve as a permanent floor.

Figure

The natural numbers escape every finite ceiling

M

\mathbb{R}

\Rightarrow

1/n

Worked examples

Example 1 (Finding an explicit $n$ for a given tolerance). Let $\varepsilon = 0.0037$ . We verify (2) concretely: find $n \in \mathbb{N}$ with $1/n < \varepsilon$ . We need $n > 1/\varepsilon = 1/0.0037 = 270.27\overline{027}$ . By the Archimedean Property (part (1), applied to $x = 1/\varepsilon$ ), such an $n$ exists; concretely we may take $n = 271$ . Check:

\frac{1}{271} = 0.0036900369\ldots < 0.0037 = \varepsilon.

Indeed $271 > 270.27027\ldots$ , confirming the bound. Note $n = 270$ fails: $1/270 = 0.0037037\ldots > 0.0037$ . This example shows the bound produced by the theorem is not merely qualitative — it pins down a specific threshold, and one must be careful about strict vs. non-strict inequalities when computing the least valid $n$ .

Example 2 (Using the two-parameter form to compare growth). Let $x = 0.02$ and $y = 500{,}000$ (so $x$ is small and positive, $y$ is large). Form (3) guarantees an $n \in \mathbb{N}$ with $nx > y$ , i.e. $0.02\,n > 500{,}000$ , i.e. $n > 25{,}000{,}000$ . Take $n = 25{,}000{,}001$ . Then

nx = 25{,}000{,}001 \times 0.02 = 500{,}000.02 > 500{,}000 = y,

as required. This illustrates the substantive content of the Archimedean Property beyond the case $x=1$ : even a minuscule step size $x = 0.02$ , repeated a finite (if large) number of times, surpasses any fixed target $y$ . This is exactly the fact used, for instance, to show a sequence like $x_n = n \cdot 0.02$ is unbounded, or that repeatedly bisecting an interval of length $0.02$ eventually produces pieces smaller than any tolerance.

Example 3 (Floor function computation, illustrating (4)). Let $x = -\pi \approx -3.14159265\ldots$ . By (4), there is a unique integer $m$ with $m \le -\pi < m+1$ . Since $-4 \le -3.14159\ldots < -3 = -4+1$ , we get $m = -4$ . Note the common error of guessing $m = -3$ : while $-3 > -\pi$ is false (in fact $-3 > -\pi$ is true since $-3 = -3.000 > -3.14159\ldots$ ), the correct containment requires $m \le x$ , and $-3 \le -\pi$ is false ( $-3 > -\pi$ ). So $m=-3$ fails the left inequality, confirming $m = -4$ is indeed the unique correct floor.

\sup \mathbb{N}

Scope, edge cases & common traps

Situation	What holds	Common mistake
$x \le 0$	$n=1$ already works: $1 > x$	Overcomplicating the proof by not splitting into cases
$x$ irrational, e.g. $x = \sqrt{2}$	Still $\exists n \in \mathbb{N}, n > \sqrt2$ (e.g. $n=2$ )	None — property is agnostic to rationality of $x$
Reciprocal form with $\varepsilon$ very large (e.g. $\varepsilon = 10^6$ )	Still true; even $n=1$ gives $1/1=1<10^6$	Assuming the theorem is only useful for tiny $\varepsilon$
Non-Archimedean ordered fields (e.g. $\mathbb{R}(t)$ , hyperreals)	Property fails; infinite/infinitesimal elements exist	Assuming every ordered field is Archimedean
Quantifier order	$\forall x\,\exists n$ — the $n$ depends on $x$	Reading it as $\exists n\,\forall x$ (a single $n$ bounding all reals) — false and would contradict unboundedness of $\mathbb{R}$ itself

The mathematician behind it

Origins · 1 minArchimedes of Syracuse and how this was discovered

Archimedes of Syracuse

c. 287 – c. 212 BC · Ancient Greek mathematics, Hellenistic Sicily

Archimedes was a mathematician, physicist, and engineer of Syracuse in Sicily, celebrated for his work on the measurement of circles, spheres, and areas under curves using an early form of exhaustion arguments — a precursor to integral calculus. The property that bears his name appears in his treatise On the Sphere and Cylinder and, more explicitly, as an axiom in Euclid's Elements (Book V, Definition 4), used to compare magnitudes of wildly different sizes by guaranteeing that a sufficiently large multiple of a smaller magnitude will exceed a larger one. Archimedes used this principle as a working tool for his method of exhaustion, allowing rigorous comparison of areas and volumes without a fully formal theory of limits — a technique not made completely rigorous until the 19th-century construction of the real numbers via completeness.

History

- c. 300 BC — Eudoxus of Cnidus develops the theory of proportions later codified in Euclid's Elements, implicitly relying on what is now called the Archimedean condition to compare ratios of magnitudes.

- c. 250 BC — Archimedes explicitly states and uses the property (sometimes called the "Axiom of Archimedes" or "Eudoxus–Archimedes axiom") in On the Sphere and Cylinder to justify his method of exhaustion for computing areas and volumes.

- 1872 — Richard Dedekind publishes his construction of the real numbers via Dedekind cuts, providing the first fully rigorous foundation from which completeness — and hence the Archimedean Property as a theorem — can be derived.

- 1899 — David Hilbert, in Grundlagen der Geometrie, studies ordered fields where the Archimedean Property fails, formalizing "non-Archimedean geometries" and clarifying that Archimedean-ness is logically independent of the other ordered-field axioms.

- 1960s — Abraham Robinson develops nonstandard analysis, constructing the hyperreal numbers

\,^*\mathbb{R}

, a non-Archimedean ordered field extension of

\mathbb{R}

containing genuine infinitesimals and infinite numbers, giving rigorous meaning to Leibniz's original infinitesimal calculus.

Is the Archimedean Property an axiom or a theorem?

In the standard development of $\mathbb{R}$ via the Completeness Axiom, the Archimedean Property is a theorem, not an independent axiom — it is derived from completeness, as shown in the proof above. Historically, however, it was treated as an axiom (Euclid's Definition V.4, and later explicit axiomatizations of ordered fields), because before Dedekind and Cantor formalized $\mathbb{R}$ via completeness, there was no deeper principle from which to derive it. In a modern rigorous course, always state it as a consequence of completeness, since completeness is the deeper and logically prior assumption, and Archimedean-ness follows for free once completeness is assumed.

Does every ordered field satisfy the Archimedean Property?

No. The property is logically independent of the ordered field axioms; it requires an extra hypothesis, and completeness is one sufficient (but not necessary) such hypothesis. The counterexample field $\mathbb{R}(t)$ of rational functions, ordered by eventual sign, is a genuine ordered field where $\mathbb{N}$ is bounded above by the element $t$ . The hyperreal numbers $\,^*\mathbb{R}$ used in nonstandard analysis are another example, containing both infinite numbers and infinitesimals. Any ordered field that is Dedekind-complete, however, is automatically Archimedean — and in fact $\mathbb{R}$ is (up to isomorphism) the unique complete ordered field, so it is the unique Archimedean-and-complete one.

Why can't we just say "obviously

n \to \infty

" without a proof?

Because " $n \to \infty$ " is an informal restatement of the very claim needing proof, and in some ordered fields it is simply false — as the $\mathbb{R}(t)$ counterexample demonstrates. Intuition trained on $\mathbb{R}$ can mislead you into thinking unboundedness of $\mathbb{N}$ is a free, structure-independent fact, but it genuinely depends on the completeness of the ambient field. The rigorous proof isolates exactly which axiom (completeness) is responsible, which is valuable both for logical hygiene and because it tells you precisely which alternative number systems can violate the property.

How does the Archimedean Property relate to the floor function

\lfloor x \rfloor

Form (4) of the theorem is exactly the existence and uniqueness of $\lfloor x \rfloor$ : the integer $m$ with $m \le x < m+1$ is by definition $\lfloor x \rfloor$ . Without the Archimedean Property, there is no guarantee that such an integer exists for every real $x$ — in a non-Archimedean field, an infinite element like $t$ in $\mathbb{R}(t)$ has no integer part at all in the usual sense, since it exceeds every integer. So the floor function, which underlies constructions like decimal expansions and the division algorithm's real-number analogues, is itself a hidden application of Archimedean-ness.

Is the Archimedean Property used to prove density of the rationals?

Yes, essentially. To show a rational number lies strictly between two reals $a < b$ , one first uses form (2) of the Archimedean Property to find $n \in \mathbb{N}$ with $1/n < b-a$ , guaranteeing "room" for a rational with denominator $n$ . Then form (4) (the floor function) locates the correct integer numerator $m = \lfloor na \rfloor + 1$ , and one checks $a < m/n < b$ . This is the standard proof of Density of the Rationals in the Reals, and it is essentially impossible to carry out without the Archimedean Property, since that is precisely what guarantees the denominator $n$ can be made large enough.

What is the intuitive difference between form (1) and form (2) of the theorem?

Form (1) talks about large numbers: no real number is an upper bound for $\mathbb{N}$ . Form (2) talks about small numbers: no positive real is a lower bound for $\{1/n : n \in \mathbb{N}\}$ other than a non-positive bound. They are two faces of the same coin, related by taking reciprocals, and the proof shows they are logically equivalent given the ordered field axioms. Intuitively, (1) says "you can't out-cap the counting numbers," while (2) says "you can't find a positive threshold that all the fractions $1/n$ stay above." Both express the absence of "infinitely large" or "infinitesimally small but positive" real numbers.

Does the Archimedean Property say anything about negative numbers or zero?

The property as stated quantifies over all $x \in \mathbb{R}$ , including negative $x$ and $x=0$ ; these cases are handled trivially since $n=1 \in \mathbb{N}$ already satisfies $1 > x$ whenever $x \le 0$ . The substantive content of the theorem is entirely about positive $x$ , especially large positive $x$ (for form (1)) or small positive $\varepsilon$ (for form (2)) — that is where unboundedness of $\mathbb{N}$ actually does work. The case split in the proof of the (2) $\Rightarrow$ (1) direction makes this explicit.