Try
Try to intersect the unit circles centered at (0,0) and (1,0), using only points with rational coordinates. Their equations give x²+y²=1 and (x−1)²+y²=1. Subtraction yields x=1/2, and then y²=3/4. Neither intersection has rational y.
In the real plane the two intersections exist. The failure over Q is not a contradiction in real Euclidean geometry; it reveals that changing the coordinate field changes existence properties. Incidence, order, congruence and continuity cannot be treated as interchangeable requirements.
Predict
Predict whether “there is always a rational number closer to √3” creates a rational square root. Approximation does not create an exact member of the field. Predict whether an ordered field must have least upper bounds. The rationals are ordered but the cut determined by t²<3 has no rational endpoint.
The Archimedean property says no positive element is larger than every positive integer. It is weaker than Dedekind completeness, the least-upper-bound property for all nonempty bounded subsets.
Experiment
Compare the requirements rather than describing all three rows as the same number line.
| Coordinate domain | Ordered field? | Every positive square root? | Least-upper-bound property? |
|---|
| Q | Yes | No | No |
| Real algebraic numbers | Yes | Yes | No |
| R | Yes | Yes | Yes |
For the middle row, positive algebraic square roots are algebraic. Yet a cut at a transcendental real number has no algebraic endpoint. This comparison uses the known existence of transcendental reals; the lesson does not prove that theorem. Explain why circle constructions can succeed while arbitrary cuts still fail.
Understand
Completeness gives positive square roots by a useful perturbation argument. Let a>0 and S be the nonnegative real t with t²≤a. It contains zero and is bounded above by a+1, so s=sup S exists.
If s²<a, choose 0<h<min(1,(a−s²)/(2s+1)). Then (s+h)²≤s²+(2s+1)h<a, contradicting the upper-bound property. If s²>a, choose 0<h<min(s,(s²−a)/(2s)). Then (s−h)²>s²−2sh>a. Every nonnegative t≥s−h is outside S, so s−h is a smaller upper bound, also a contradiction. Therefore s²=a.
Completeness also implies the Archimedean property. If the positive integers were bounded, their supremum s would exist. Since s−1 is not an upper bound, some integer n>s−1; then n+1>s, contradiction. These arguments identify the exact role of supremum existence rather than appealing to a smooth-looking line.
Reverse recovery after the synthetic coordinate lemmas
The previous model lesson goes from R² to geometry. Here is the reverse implication with its assumptions exposed. This is a reconstruction theorem from a coordinate-lemma package, not a compressed proof that all those lemmas follow from an arbitrary axiom list.
Assume a synthetic Euclidean metric plane has two perpendicular oriented axes with a common origin O and the following proved or stipulated facts:
- Each axis has a real ruler, with O at 0 and distances the absolute differences of ruler numbers.
- Parallel projection onto the axes associates to every point P one pair (x(P),y(P)), and for every real pair there is exactly one point with those projections. This is the rectangle-existence and uniqueness lemma.
- For two points P,Q, the parallel rectangle through them has perpendicular leg lengths |x(P)−x(Q)| and |y(P)−y(Q)|. This is the parallel-transport/rectangle congruence lemma, not merely a drawing assumption.
- The Pythagorean theorem holds for these right triangles, and points on a full line have a real ruler with metric betweenness: B lies between A,C exactly when d(A,B)+d(B,C)=d(A,C).
These are the explicit Euclidean ingredients one must derive from a chosen synthetic foundation before invoking this result. In particular, the rectangle lemmas use a Euclidean parallel principle; they are not neutral consequences. They do not hold for the hyperbolic plane with its hyperbolic metric.
Define Φ(P)=(x(P),y(P)). The rectangle existence/uniqueness clause makes Φ a bijection onto R². Apply Pythagoras to its legs to obtain
d(P,Q)²=(x(P)−x(Q))²+(y(P)−y(Q))².
The zero-leg cases reduce directly to the axis/parallel-ruler statement. Thus Φ preserves distances. It preserves metric betweenness because the distance-addition equality is preserved; in the real normed plane equality in the triangle inequality characterizes points on the same directed segment, as established in the coordinate-model lesson.
It also carries complete synthetic lines to complete real affine lines. Fix distinct A,B on a synthetic line. Any point C on that line is between them or beyond an endpoint, so its image is collinear with Φ(A),Φ(B) by the preserved metric equalities. Conversely the ruler gives a synthetic point at every signed distance from A along either ray. Its image is the unique collinear point at that signed distance; hence the entire real affine line is reached, not merely a subset. Distance-preserving bijections and their inverses are continuous, so this is simultaneously a metric and topological identification. Congruence is preserved by its distance definition; angle values are recovered from side lengths through the cosine law once the synthetic angle/triangle theory supplies that relation.
Two coordinate choices satisfying this package differ by a Euclidean isometry: their composition is a distance-preserving bijection of R², classified in the orthogonal-map lesson. The origin and orientation choices change coordinates, not the geometry. This is a genuine reverse conclusion, with every potentially circular premise listed before the proof.
Test the boundary rather than pretending categoricity
Try the same argument over the rational plane. Which step fails if every real ruler coordinate is required? Which step fails for hyperbolic coordinates even though they use real numbers? Does proving the reconstruction from this package automatically derive the package from the neutral axioms?
Feedback and rubric
The rational plane cannot supply a point for every real ruler number or every real pair, so the first two assumptions already fail. Hyperbolic coordinates may be real and bijective as a set parametrization, but the rectangular distance/Pythagorean package fails: its metric is not the Euclidean sum of squared coordinate differences. The final question's answer is no. Deriving the package from a specific complete Hilbert or Birkhoff system is a further synthetic theorem; the reconstruction proof identifies exactly what that derivation must establish and cannot be used to assume its conclusion. A complete answer distinguishes field completeness, Euclidean metric structure and the logical direction of a theorem.Apply
Repair “The rational plane satisfies the entire Hilbert system because its lines look normal.” Besides continuity, some congruence transport axioms may fail when a required length needs an absent square root. A model claim must inspect every chosen axiom, not just line incidence.
For optional depth, full synthetic coordinatization is a substantial theorem: under appropriate incidence, order, congruence and continuity assumptions, coordinates recover the real Euclidean plane. We use that as a named structural boundary, not as a theorem proved by one circle calculation. Different continuity packages can produce different classes of coordinate fields.
Continue
For each construction, ask which existence principle produces the new point. An algebraic formula with a square root still needs that root to belong to the domain. Continue to neutral triangle theorems to trace results that do not require a Euclidean parallel postulate.
