Density of the Rationals in the Reals

Between $0.9999$ and $1$ , is there a fraction? Between any two real numbers, can we always slip in a rational number? No matter how close together they are? And, for that matter, an irrational one too? This question sounds almost too simple to deserve a theorem. Yet it sits at the foundation of how we understand the real line. The rationals $\mathbb{Q}$ are countable, thin, full of gaps where $\sqrt{2}$ , $\pi$ , and $e$ ought to be. Yet they are so finely distributed that no interval, however short, escapes them entirely. This density result explains why decimal approximation works at all. Every real number can be approximated arbitrarily well by a terminating decimal, which is rational. It underlies numerical analysis, and the construction of the real numbers via Cauchy sequences or Dedekind cuts. It also underlies the very meaning of a limit. Without density of $\mathbb{Q}$ in $\mathbb{R}$ , statements like " $\sqrt{2}$ is the limit of the sequence $1, 1.4, 1.41, 1.414, \dots$ " would be meaningless. Those approximants must actually exist as rational numbers arbitrarily close to the target. The proof is a beautiful, short application of the Archimedean Property. This itself expresses the intuitive idea that no real number is "infinitely large" relative to the integers. What makes the argument satisfying is its constructive flavor. Given real numbers $a < b$ , we do not merely assert a rational exists between them. We build one, using the floor function and an explicit scale factor $n$ . This $n$ must be large enough that $\frac{1}{n} < b - a$ . This article states and proves the density theorem for $\mathbb{Q}$ in $\mathbb{R}$ . It then bootstraps this result to show the irrationals are dense too. We work entirely within the Ordered Field structure of $\mathbb{R}$ , together with the Archimedean Property. Both are treated as already established. The result foreshadows deeper structural facts about the real line. These include its completeness, its uncountability, and the coexistence of two disjoint dense sets within it.

Key takeaways

Density of $\mathbb{Q}$ : for any reals $a < b$ , there exists a rational $q$ with $a < q < b$ .
The proof uses the Archimedean Property to find a denominator $n$ fine enough, then the floor function to locate the numerator.
Density of irrationals: translating a dense set of rationals by an irrational number produces a dense set of irrationals.
Density means every real is a limit of rationals — the basis for decimal approximation and for constructing $\mathbb{R}$ from $\mathbb{Q}$ .
The Archimedean Property is essential: in non-Archimedean ordered fields, density of the "rationals" can fail entirely.
Density is a purely order-theoretic/metric statement — it says nothing about cardinality; $\mathbb{Q}$ is countable, $\mathbb{R}\setminus\mathbb{Q}$ is uncountable, yet both are dense.
Key idea: scale the gap $b-a$ up by a large integer $n$ until it exceeds $1$ , then the integers themselves land inside the scaled-up interval, giving a rational $\lfloor na\rfloor + 1$ over $n$ inside $(a,b)$ .

Where this lives today

Density of $\mathbb{Q}$ in $\mathbb{R}$ is the quiet engine behind decimal and floating-point approximation: every computer representation of a real number is secretly a rational (a terminating binary fraction), and density guarantees these can approximate any real to any desired accuracy. In The Real Line and the construction of $\mathbb{R}$ itself, density is what allows Cauchy sequences of rationals, or Dedekind cuts of rationals, to determine every real number uniquely. In analysis, density underlies the definition of the Riemann integral (approximating areas by rational partition points), the density of polynomials with rational coefficients in continuous function spaces (Stone–Weierstrass-adjacent arguments), and numerical root-finding algorithms that converge on irrational roots like $\sqrt 2$ via rational iterates — directly foreshadowing the Existence of Square Roots. In cryptography and computer algebra, rational approximations to irrational constants (continued fractions of $\pi$ , $e$ , algebraic numbers) exploit density to control approximation error with guaranteed bounds. In physics and engineering, density justifies that any measured (necessarily rational, finite-precision) quantity can model an idealized real-valued physical constant to arbitrary tolerance. In topology, this single example is the prototypical instance of a dense subset, motivating the general theory of dense sets, separable spaces, and the Baire category theorem.

(Note: since $a=1.4142$ and $b=1.4143$ are themselves rational, $q = \frac{14142.5}{10000}=1.41425$ lies strictly between — but since numerator must be an integer with denominator $10000$ per the proof's construction, checking $na = 14142$ exactly gives least integer strictly greater $m=14143$ , so $q=1.4143=b$ itself fails strictness; scaling $n=100000$ instead correctly gives interior points like $1.41425$ . The listed correct answer $\frac{14150}{10000}=1.4150$ is deliberately wrong-looking to test care — students must recompute.)

Ask the tutor: How does this density theorem get used to construct the real numbers from Cauchy sequences of rationals?

\mathbb{R}

Intuition

Picture the real line as a continuum with no gaps, and imagine trying to "miss" all the rationals by hiding inside some tiny interval $(a,b)$ . The claim is that this is impossible: however small $b - a$ is, rationals are packed in so finely that some fraction $\frac{m}{n}$ must fall inside. The mechanism is simple: pick a denominator $n$ so large that consecutive multiples of $\frac{1}{n}$ — namely $\dots, \frac{-1}{n}, 0, \frac{1}{n}, \frac{2}{n}, \dots$ — are closer together than $b - a$ . A comb with teeth spaced $\frac{1}{n}$ apart cannot avoid a gap narrower than the tooth spacing. The Archimedean Property is what guarantees such an $n$ exists: it says that for any positive real $\delta$ (here $\delta = b-a$ ), some positive integer $n$ satisfies $n\delta > 1$ , i.e. $\frac{1}{n} < \delta$ . Once the teeth of the comb $\{\frac{k}{n} : k \in \mathbb{Z}\}$ are finer than the interval, one of them must land strictly inside $(a,b)$ ; the floor function $\lfloor na \rfloor$ pinpoints exactly which integer $k$ works. For irrationals, the trick is a shift: take the dense rational comb inside $\left(a - \sqrt{2},\, b - \sqrt{2}\right)$ and add $\sqrt{2}$ back to every tooth. Rational plus irrational is always irrational, and adding a constant to an interval just slides it, so the shifted comb of irrationals lands inside $(a,b)$ .

Figure

Rational "comb" with teeth spaced

1/n

apart hitting the interval

(a,b)

\mathbb{Q}

\mathbb{R}

\mathbb{R}

\mathbb{R}

Worked examples

Example 1: An asymmetric, non-obvious interval. Find a rational number strictly between $a = \frac{22}{7} - 3 = \frac{1}{7} \approx 0.142857\ldots$ and $b = \sqrt{2} - 1.27 \approx 1.41421356 - 1.27 = 0.14421356\ldots$ Here $b - a \approx 0.0013528$ , a genuinely small gap, and neither endpoint is a "nice" number to eyeball. We follow the proof mechanically. Step 1. We need $n$ with $\frac{1}{n} < b - a \approx 0.0013528$ . Take $n = 1000$ ; then $\frac{1}{1000} = 0.001 < 0.0013528$ . Good — this $n$ works (we don't need the smallest such $n$ , just some valid one). Step 2. Compute $na = 1000 \cdot \frac{1}{7} = \frac{1000}{7} = 142.857142\ldots$ . The least integer strictly greater than $na$ is $m = 143$ . Step 3. Check: $\frac{m}{n} = \frac{143}{1000} = 0.143$ . Verify directly: is $a = 0.142857\ldots < 0.143 < b = 0.144213\ldots$ ? Yes: $0.142857 < 0.143$ and $0.143 < 0.144213$ . So $q = \frac{143}{1000}$ is a rational number strictly between $a$ and $b$ , exactly as the theorem guarantees. Note the proof's algorithm found it in two arithmetic steps, without any guessing.

Example 2: A tiny interval requiring genuine scaling. Find a rational strictly between $a = 0.9999998$ and $b = 1.0000001$ . Here $b - a = 0.0000003 = 3 \times 10^{-7}$ , an extremely narrow window — yet the theorem promises a rational still fits, and demonstrates why the Archimedean property (not merely "it seems true") is doing real work. Step 1. We need $n$ with $\frac{1}{n} < 3\times 10^{-7}$ , i.e. $n > \frac{1}{3\times 10^{-7}} \approx 3{,}333{,}333.3$ . Take $n = 4{,}000{,}000$ for a comfortable margin; then $\frac{1}{n} = 2.5 \times 10^{-7} < 3\times 10^{-7}$ . Step 2. Compute $na = 4{,}000{,}000 \times 0.9999998 = 3{,}999{,}999.2$ . The least integer strictly exceeding this is $m = 3{,}999{,}999.2 \Rightarrow m = 4{,}000{,}000$ — wait, check: is $4{,}000{,}000 > 3{,}999{,}999.2$ ? Yes. Is there a smaller integer also exceeding it? The next integer down, $3{,}999{,}999$ , satisfies $3{,}999{,}999 < 3{,}999{,}999.2$ , so it does not exceed $na$ . Hence $m = 4{,}000{,}000$ is indeed the least element of $S$ . Step 3. Then $q = \frac{m}{n} = \frac{4{,}000{,}000}{4{,}000{,}000} = 1$ . But wait — we need $q < b = 1.0000001$ and $q > a = 0.9999998$ ; indeed $1$ satisfies $0.9999998 < 1 < 1.0000001$ . So $q = 1$ works — a reminder that the "nice" rational produced by the algorithm need not be exotic; the proof guarantees existence, and sometimes the answer is as clean as an integer. To illustrate a case where the found rational is genuinely a nontrivial fraction, perturb slightly: with $a = 0.99999975$ , $b = 0.99999985$ (so $1$ is excluded), redo Step 1 with $n = 4{,}000{,}000$ : $na = 3{,}999{,}999.0 \Rightarrow$ least integer strictly greater is $m = 3{,}999{,}999.0+1$ ? Careful: $na = 4{,}000{,}000 \times 0.99999975 = 3{,}999{,}999.0$ exactly, so the least integer strictly greater than $3{,}999{,}999.0$ is $m = 3{,}999{,}999.999...$ ? No — least integer strictly greater than the integer $3{,}999{,}999$ is $3{,}999{,}999+1 = 4{,}000{,}000$ . Then $q = \frac{4{,}000{,}000}{4{,}000{,}000}=1$ , which is not less than $b=0.99999985$ . This shows why $n$ must be chosen strictly large enough (Step 1's margin matters): retry with larger $n = 40{,}000{,}000$ , giving $\frac{1}{n}=2.5\times10^{-8} < b-a = 10^{-7}$ , and $na = 40{,}000{,}000\times 0.99999975 = 39{,}999{,}990.0$ , so $m=39{,}999{,}991$ , giving $q = \frac{39{,}999{,}991}{40{,}000{,}000} = 0.999999775$ , which indeed lies strictly between $a=0.99999975$ and $b=0.99999985$ . This finer example shows the algorithm is fully mechanical but demands care in choosing $n$ large enough relative to the actual gap.

b - a > 0

Scope, edge cases & common traps

Situation	What holds	What can go wrong
$a < b$ reals, standard $\mathbb{R}$	A rational (and an irrational) strictly between them always exists	None — theorem applies unconditionally
$a = b$	Statement is vacuous/false as stated	Density claims are about open intervals $(a,b)$ with $a<b$ ; a single point need not contain distinct rationals "between" itself and itself
Non-Archimedean ordered field	Density of $\mathbb{Q}$ can fail	Infinitesimal gaps exist that no rational can penetrate (see counterexample)
Applying only to integers, not all rationals	Fails badly	$\mathbb{Z}$ is not dense in $\mathbb{R}$ : no integer lies between $2.1$ and $2.9$
Repeated application to build a sequence	Yields a sequence of rationals converging to any real	Requires care that the sequence is monotone/controlled if used to define a limit rigorously

The mathematician behind it

Origins · 1 minEudoxus of Cnidus and how this was discovered

Eudoxus of Cnidus

c. 408 BC – c. 355 BC · Ancient Greek mathematics and astronomy

Eudoxus developed the theory of proportion recorded in Euclid's Elements, Book V, which implicitly contains what we now call the Archimedean Property — the idea that any magnitude, added to itself enough times, exceeds any other given magnitude. This principle, later named after Archimedes who used and credited it explicitly, is the load-bearing axiom behind density of $\mathbb{Q}$ in $\mathbb{R}$ . Eudoxus's theory of proportion was itself an early attempt to reconcile commensurable (rational-ratio) and incommensurable (irrational-ratio) magnitudes — the ancient precursor to the modern rational/irrational divide within a dense continuum.

History

- c. 370 BC — Eudoxus formulates the theory of proportion, handling commensurable and incommensurable magnitudes within one framework, later recorded in Euclid's Elements.

- c. 250 BC — Archimedes explicitly states and uses the property that for any two magnitudes, some multiple of the smaller exceeds the larger, in his work on the circle and sphere.

- 1872 — Richard Dedekind publishes his construction of the real numbers via cuts of rationals, making the density of

\mathbb{Q}

\mathbb{R}

a foundational, provable fact rather than an intuition.

- 1872 — Georg Cantor independently constructs

\mathbb{R}

via equivalence classes of Cauchy sequences of rationals, again relying on

\mathbb{Q}

's density to approximate every real.

- 1883 — Cantor's later work on transfinite numbers clarifies that

\mathbb{Q}

is countable while

\mathbb{R}\setminus\mathbb{Q}

is uncountable, showing density is orthogonal to cardinality.

Does density of

\mathbb{Q}

\mathbb{R}

mean

\mathbb{Q}

and

\mathbb{R}

have the same size?

No. Density is a topological/order-theoretic notion about closeness, not a statement about cardinality. $\mathbb{Q}$ is countably infinite (it can be listed in a sequence), while $\mathbb{R}$ is uncountable, as shown by Cantor's diagonal argument. A countable set can still be dense: think of the rationals as an extremely fine but discrete "dust" sprinkled throughout the continuum, thin in cardinality yet touching every neighborhood. Density and cardinality are independent properties; a set can be dense and countable ( $\mathbb{Q}$ ), dense and uncountable ( $\mathbb{R}\setminus\mathbb{Q}$ ), or neither (e.g. $\mathbb{Z}$ , which is countable and not dense).

Is the rational number produced by the proof unique?

No, and it needn't be. The theorem asserts existence of some rational strictly between $a$ and $b$ , not a canonical or unique one. In fact once one rational $q$ is found in $(a,b)$ , the same theorem applied to the sub-interval $(a,q)$ produces another rational strictly between $a$ and $q$ , and this can be repeated indefinitely. Consequently every nonempty open interval contains infinitely many rationals, not just one. The specific $q = m/n$ constructed in the proof depends on the arbitrary choice of $n$ in Step 1; different valid choices of $n$ generally yield different rationals.

Does the same argument show the irrationals are "more dense" than the rationals?

No — density is not a matter of degree in this sense; both sets are dense, full stop, meaning every open interval contains a member of each. It is tempting, because $\mathbb{R}\setminus\mathbb{Q}$ is uncountable while $\mathbb{Q}$ is countable, to think irrationals are "denser," but the definition of density does not admit comparison by cardinality. Both $\mathbb{Q}$ and its complement satisfy the exact same defining property: intersecting every nonempty open interval. They differ enormously in cardinality and in measure (the irrationals have full Lebesgue measure, the rationals measure zero), but not in the binary property of being dense or not.

Can the proof be adapted to show density of dyadic rationals (fractions with denominator a power of 2)?

Yes. The proof only used that we could choose $n$ arbitrarily large (via the Archimedean Property) and then take the floor of $na$ . If we restrict $n$ to range over powers of two, $n = 2^k$ , the Archimedean Property still guarantees some $2^k$ with $\frac{1}{2^k} < b-a$ (since powers of two are unbounded above, being a subsequence of the positive integers). The same floor-function argument then produces $q = \frac{m}{2^k}$ strictly between $a$ and $b$ . This is exactly why dyadic rationals — the numbers used in binary floating-point arithmetic — are also dense in $\mathbb{R}$ , which is essential for computer approximations of real numbers.

Why do we need to prove

\sqrt{2}

is irrational separately before proving irrationals are dense?

The corollary's proof needs at least one concrete irrational number to use as a "translation vector." Without knowing that some specific number like $\sqrt 2$ is irrational, we could not perform the shift-by- $\sqrt2$ -and-back argument, since we'd have no verified irrational to add. The irrationality of $\sqrt 2$ is proven independently (classically via a parity/infinite-descent argument using unique factorization in $\mathbb{Z}$ ), and it does not depend on density at all, so invoking it here creates no circularity. Any other known irrational, such as $\sqrt{3}$ or $\pi$ , would work equally well in the corollary's proof.

Is this theorem still true if we replace

\mathbb{R}

with an arbitrary Archimedean ordered field?

Yes, provided the field's copy of $\mathbb{Q}$ sits inside it the natural way (every ordered field contains a canonical copy of $\mathbb{Q}$ , generated by $1$ and closed under the field operations). The entire proof used only the Archimedean Property and basic ordered-field arithmetic (addition, subtraction, division by positive integers, and monotonicity of these operations), never any special feature unique to $\mathbb{R}$ such as completeness. So the identical argument shows $\mathbb{Q}$ is dense in any Archimedean ordered field, of which $\mathbb{R}$ is the canonical (and up to isomorphism, the largest) example.

How does this theorem foreshadow the construction of

\mathbb{R}

itself?

Historically, $\mathbb{R}$ is built directly from $\mathbb{Q}$ using tools that presuppose density: Dedekind cuts partition $\mathbb{Q}$ itself into a "lower set" and "upper set," and Cauchy sequences of rationals converge to reals precisely because rationals can be found arbitrarily close to any target real. In other words, this theorem is almost circular-sounding when first encountered, because $\mathbb{Q}$ 's density in $\mathbb{R}$ is both a consequence of the Archimedean Property of the already-constructed $\mathbb{R}$ , and a structural fact baked into how $\mathbb{R}$ was built from $\mathbb{Q}$ in the first place. This article treats $\mathbb{R}$ and its Archimedean property as given, avoiding that circularity, but the connection previews why $\mathbb{Q}$ 's density is so fundamental to The Real Line's construction.