The single most counter-intuitive fact about the number line, and the one you want in your head as a one-line mental model: the rationals are everywhere, and yet they are full of holes. They are like a sieve — densely woven, but with gaps that \sqrt{2} and \pi and e slip right through. Once this picture clicks, a huge chunk of analysis stops feeling mysterious.
Densely everywhere
Take any two distinct rationals, no matter how close. Say \dfrac{1}{1000000} and \dfrac{1}{999999}. Is there another rational between them? Yes — their average:
is a rational (sum and quotient of rationals), and it sits exactly in the middle.
Why the average works: a rational plus a rational is a rational, and a rational divided by 2 is a rational. So the midpoint of two rationals is always rational. Between any two rationals, there is always a third — and between those two, a fourth, and so on forever.
This property is called density. No matter how close together two rationals sit, you can always wedge infinitely many more between them. Zoom in as far as you like; rationals keep appearing.
So, in what sense could the rationals possibly be missing anything? If they are packed in so tightly that any neighbourhood contains infinitely many, what room is left for \sqrt{2} to live in?
Yet riddled with holes
Here is the shock. Take \sqrt{2} \approx 1.41421356\dots. This is a real, measurable length — the diagonal of a unit square. It sits at one specific point on the number line. And that point is not a rational. The proof is the classical one: assume \sqrt{2} = \dfrac{p}{q} in lowest terms, square both sides, derive that p and q are both even — contradicting "lowest terms."
So the rational at position \sqrt{2} on the number line is a hole. The rationals cluster arbitrarily close to it on both sides, but none of them land on it.
And this is not an isolated incident. There is a hole at every irrational: at \pi, at e, at \sqrt{3}, at \log_2 5, at every root of an integer that isn't a perfect power. In fact, as a later satellite shows, the irrationals outnumber the rationals so overwhelmingly that "almost every" real is a hole from the rationals' perspective.
The sieve image
The mental model to carry is this: the rationals form a sieve with infinitely fine mesh, and the irrationals are the grains of sand that slip through the holes.
The sieve image resolves the paradox. "Densely everywhere" and "full of holes" sound contradictory if you are thinking of density in the everyday sense — like a crowd of people with no gaps. But mathematical density is different. It says: given any gap, I can shrink it as much as you want. It does not say there are no gaps at all. The sieve can have infinitely fine mesh and still leave infinitely many points uncovered.
Why the distinction matters: two sets can both be dense and still be disjoint. The rationals are dense in \mathbb{R}. The irrationals are also dense in \mathbb{R} — between any two rationals sits an irrational. Density is about getting close, not about filling up.
Where the mental model pays off
Carry the sieve image and suddenly several facts stop feeling strange:
-
Why \sqrt{2} has a decimal expansion that never terminates and never repeats. A terminating or repeating decimal is rational — a tick on the sieve. \sqrt{2} is between ticks, so its decimal has to keep running forever, never locking in to a repeating block.
-
Why you can approximate any irrational to arbitrary accuracy by a rational. The sieve mesh is infinitely fine — so for any \epsilon > 0 you want, there is a rational within \epsilon of \sqrt{2}. That is the basis of all numerical computing. Your calculator displays 1.41421356 because that rational is close enough to the hole.
-
Why "close enough" is never "equal." No matter how many decimal digits you compute, the rational approximation is still a tick, not the hole itself. You can chase \sqrt{2} forever and never land on it with a fraction.
-
Why the real numbers are needed for calculus. Limits require completeness — the property that every Cauchy sequence has a limit in the set. The rationals do not have this: the sequence 1, 1.4, 1.41, 1.414, \dots is Cauchy but its limit \sqrt{2} is not rational. Calculus is done in \mathbb{R} because \mathbb{R} has no holes.
A quick exam-time use
On a JEE question, if you see a phrase like "the sequence of rationals \{a_n\} converges to L" and it is not immediately clear whether L is rational, default to assume it could be irrational. The rationals do not have to converge to another rational. They can converge to anything on the number line — because every real number, rational or irrational, has rationals clustered arbitrarily close to it.
This is the mental model behind half of the properties of real numbers: the rationals are dense in \mathbb{R}, the irrationals are dense in \mathbb{R}, the whole real line is one long tapestry where the two kinds of number live interleaved. The sieve image captures it in one line. File it.
The sieve in action: squeezing √2
You want to pin down \sqrt{2} using only rationals. Start with the interval [1, 2] — both endpoints rational. You know \sqrt{2} is inside because 1^2 < 2 < 2^2. Now bisect:
- Midpoint: 1.5. Is 1.5^2 = 2.25 > 2? Yes. So \sqrt{2} \in [1, 1.5].
- Midpoint: 1.25. Is 1.25^2 = 1.5625 < 2? Yes. So \sqrt{2} \in [1.25, 1.5].
- Midpoint: 1.375. Is 1.375^2 = 1.890625 < 2? Yes. So \sqrt{2} \in [1.375, 1.5].
- Midpoint: 1.4375. Is 1.4375^2 = 2.06640625 > 2? Yes. So \sqrt{2} \in [1.375, 1.4375].
After four steps, the interval width is 0.0625. After twenty steps, it is \approx 10^{-6}. Each endpoint is rational; \sqrt{2} is always strictly between two rational endpoints. The sieve keeps shrinking around the hole, but the hole itself never becomes a tick. That is what every dense-but-not-complete statement looks like in practice.
Related: Number Systems · Are Irrationals Rare? They Outnumber Rationals · Zooming the Number Line: Rationals Density · If Rationals Are Dense, How Do Irrationals Fit In?