Dedekind Cut Visualiser: a Real Number IS a Split of ℚ

Most students meet \sqrt{2} as a symbol — a thing you cannot write as p/q — and then the textbook moves on. But \sqrt{2} is not a symbol. It is a location, and the cleanest way to say where that location is, without secretly assuming the real numbers already exist, is Dedekind's trick: describe the number by where it splits the rationals.

This satellite is the trick, made into a toy you can drag.

The idea in one paragraph

A Dedekind cut is a way of slicing the rational number line \mathbb{Q} into two pieces: a lower set L (everything below the cut) and an upper set U (everything above). Every rational has to end up in exactly one of the two. A real number is such a pair (L, U). If the cut lands on a rational like 3/2, then 3/2 is either the largest element of L or the smallest of U — one of the two sides owns it. If the cut lands on an irrational like \sqrt{2}, something strange happens: L has no largest element, U has no smallest, and the cut lives in the gap between them. That gap is the real number.

The widget

cut:

Blue dots: rationals in the lower set $L$. Orange dots: rationals in the upper set $U$. Drag the cut. At $\sqrt{2}$, neither side has a boundary element — the blue dots creep up but never touch, and the orange dots creep down but never touch. That untouchable gap is the real number $\sqrt{2}$.

The blue dots on the left half are the rationals that fall into the lower set L. Their sizes are bigger when the denominator is smaller — so 1/2, 3/2, 2 are large dots, and things like 17/20 are tiny. The orange dots on the right are the rationals that fall into the upper set U, coloured the same way. The vertical red line is the cut: the position at which the rationals get split in two.

Grab the slider and try the three buttons in turn.

Snap to 3/2. The cut lands exactly on a rational dot, coloured red. Every rational below 3/2 is blue; every rational above is orange. The rational 3/2 itself has to go somewhere — by convention, we put it in U (so U has a smallest element, namely 3/2). In this case L has no largest element — because whatever rational you name below 3/2, there is always a bigger one still below (take its average with 3/2, for instance). So L has no max, but U has a min, and the min is the cut. Flip the convention and the opposite holds. Either way, one of the two sides owns the boundary.

Snap to \sqrt{2}. Now something different happens. The cut sits in a gap between blue and orange dots. Zoom in with your eye: the blue dots crowd upward toward the cut but never touch it; the orange dots crowd downward but never touch it. There is no blue dot "nearest to the cut," and no orange dot "nearest to the cut." In symbols:

L = \{q \in \mathbb{Q} : q < 0 \text{ or } q^2 < 2\} has no largest element
U = \{q \in \mathbb{Q} : q > 0 \text{ and } q^2 > 2\} has no smallest element

Snap to \pi. Same story as \sqrt{2} — L has no max, U has no min, and \pi lives in the untouched gap.

Why L has no largest element at \sqrt{2}

Suppose, for contradiction, that L has a largest element m. Then m is rational, m < \sqrt{2}, and no rational is between m and \sqrt{2}. But that contradicts the density of the rationals: between any two real numbers you can always find a rational. In particular, the rationals are dense enough that between m and \sqrt{2} there is a rational m' > m with m'^2 < 2, so m' \in L and m' > m — contradicting that m was the largest. See the companion Rationals Are Dense in the Reals for the explicit "average" construction.

Why this argument works only at irrational cuts: the cut point itself has to be missing from \mathbb{Q}. If the cut were rational, it would sit as the smallest element of U (or largest of L, depending on convention), and the density argument would still find no element of L between m and the cut — but that's fine, because the cut itself is in U, not in the gap. Density kills "largest element of L" only when the cut is outside \mathbb{Q} entirely.

The cut IS the number

Here is where Dedekind's idea stops looking like a gimmick and starts looking like a definition.

We do not yet have \mathbb{R}. We only have \mathbb{Q}, plus the ability to talk about subsets of \mathbb{Q}. Dedekind says: a real number is a partition (L, U) of \mathbb{Q} into two non-empty sets such that

every rational is in exactly one of L or U,
every element of L is less than every element of U,
L has no largest element.

Rule 3 is a convention-fixing move: it forces the cut-point to go to U whenever there is one. Then:

Cuts where U has a smallest element correspond to rational numbers (the smallest of U is the rational).
Cuts where U has no smallest element correspond to irrational numbers (the cut lives in a gap).

Every real number you have ever used is one of these two kinds of cut. \sqrt{2} is the cut described above. \pi is the cut where L is the set of rationals less than the circumference-to-diameter ratio, defined by squeezing from below with polygons. e is the cut defined by the partial sums of \sum 1/n!.

This is not just a clever restatement. It is a construction of \mathbb{R} from \mathbb{Q} alone, and it was the first time in history anyone had built the real numbers without hand-waving. Before Dedekind (1872), the reals were described geometrically — "a continuous line" — and the circular arguments went undetected for centuries.

Why this construction gives completeness for free

The payoff: once you define real numbers as cuts, completeness (every bounded set of reals has a supremum) is almost automatic.

Proof sketch. Let S be a non-empty set of real numbers bounded above. Each real in S is a cut (L_\alpha, U_\alpha). Form the union L = \bigcup L_\alpha and its complement U in \mathbb{Q}. This pair (L, U) is itself a Dedekind cut, and it is the least cut bigger than every element of S — that is, the supremum. Because it is a cut, it is a real number. No extra axiom needed; the construction builds supremum-existence into \mathbb{R} from the start.

This is why textbooks say "the reals are complete by construction." The Dedekind cut definition is not the only way to build \mathbb{R} (Cauchy sequences of rationals give another), but it is the most geometric: the picture in the widget — a cut splits the rationals and the cut itself is the new number — is the definition itself, animated.

What to take away

Stop thinking of irrationals as "weird numbers with infinite decimals." Start thinking of them as locations on the rational line that no rational occupies. Every irrational is a gap between the rationals that squeeze it from below and the rationals that squeeze it from above. The gap has a name (\sqrt{2}, \pi, e) and a behaviour (the sets L and U bordering it have no max and no min respectively), and that pair of facts — the behaviour plus the surroundings — is the number.

Once you see the reals as cuts, three things become clear at once:

Why decimal expansions of irrationals don't repeat. A repeating expansion would pin the cut to a rational location (geometric series converge to rationals). A non-repeating expansion describes a cut with no rational cut-point. The non-repetition is the cut-in-a-gap-ness, in digit form.
Why the rationals are dense but not complete. Density is what makes the cut well-defined — there are rationals arbitrarily close to it on both sides. Incompleteness of \mathbb{Q} is what makes the cut necessary — the cut point itself is not a rational.
Why calculus needs \mathbb{R}, not \mathbb{Q}. Every limit, every supremum-based argument, every intermediate-value application depends on cuts existing as numbers. On \mathbb{Q} alone, you can describe the cut but cannot name it. On \mathbb{R}, every cut has a name — and that is what a real number is.