Coordinatization and the Theory of $\mathbb{R}$

The independence proof closed the logical question about the parallel postulate using models. This page pushes the model-theoretic viewpoint to its sharpest conclusion: elementary Euclidean geometry and the first-order theory of the real numbers are the same object. Descartes' analytic geometry made points into coordinate pairs; Hilbert showed the coordinate field can be built from the synthetic axioms themselves; and Tarski proved the resulting theory — the theory of real closed fields — is complete and decidable. The upshot is a dictionary in which geometric figures are semialgebraic sets, geometric truth is polynomial algebra, and Euclid's Elements is, in principle, a decidable theory. This page is the bridge from the synthetic and logical half of the spine to the coordinate/metric half, and it deepens the model-theory connection sketched in independence.md §5.

References: Hilbert, Grundlagen der Geometrie, chs. 3, 5–6 (segment arithmetic); Tarski, A Decision Method for Elementary Algebra and Geometry (1948/1951); Schwabhäuser–Szmielew–Tarski, Metamathematische Methoden in der Geometrie; Marker, Model Theory: An Introduction, ch. 3; Bochnak–Coste–Roy, Real Algebraic Geometry.

1. Coordinatization: Building the Field from the Axioms

Descartes (1637) supplied the informal bridge — a point is a pair $(x,y)$ — but the substantive result is Hilbert's segment arithmetic (Streckenrechnung): starting from the synthetic axioms alone, with no numbers assumed, one constructs an ordered field and a bijection of the plane onto its square. We first fix the target notion (the same one used in analysis/real-numbers.md):

Definition (ordered field). A field is a set $F$ equipped with two binary operations $+$ and $\cdot$ and two distinguished elements $0\ne 1$ such that

• $(F,+,0)$ is an abelian group (associative, commutative, identity $0$, inverses $-x$);
• $(F\setminus \{0\},\cdot ,1)$ is an abelian group (associative, commutative, identity $1$, inverses $x^{-1}$);
• multiplication distributes over addition: $x(y+z)=xy+xz$.

An ordered field is a field $F$ together with a total order $<$ compatible with the operations: for all $x,y,z\in F$, $$\text{(O1)}x<y⟹x+z<y+z,\text{(O2)}x>0\text{ and }y>0⟹xy>0.$$

1.1 The construction

Construction (Hilbert's segment arithmetic). Fix a line $\ell$ and two distinct points $O,U\in \ell$.

• Carrier. Let $F$ be the set of points of $\ell$. Set $0:=O$ and $1:=U$. (Equivalently, a point $P\in \ell$ names the congruence class of the directed segment $\overrightarrow{OP}$, so $F$ is the set of signed segment-lengths along $\ell$.)
• Addition $+:F\times F\to F$. For $a,b\in F$, let $a+b$ be the unique point of $\ell$ obtained by laying off the directed segment $\overrightarrow{Ob}$ with its tail at $a$ (rigid transport, guaranteed by the congruence axioms III). The additive inverse $-a$ is the reflection of $a$ across $O$.
• Multiplication $\cdot :F\times F\to F$. Fix an auxiliary line $m$ through $O$ with $m\ne \ell$. For $a,b\in F$, define $a\cdot b$ to be the unique point $c\in \ell$ given by the intercept theorem — $c$ is characterized by the proportion $$Oc:Ob=Oa:OU,$$ constructed by transferring $a$ to $m$ and drawing the parallel that cuts off $c$ on $\ell$; the required parallels exist by the parallel axiom.
• Order $<$. Declare $a<b$ iff $b$ lies on the ray from $O$ through $U$ strictly beyond $a$ — i.e. the betweenness relation of the order axioms II, restricted to $\ell$.

Theorem (Hilbert, Grundlagen §§13–15). With these operations $(F,+,\cdot ,0,1,<)$ is an ordered field. Moreover: $$\text{Pappus}⟺\text{multiplication is commutative},\text{Desargues}⟺\text{(F,+,⋅) is an (associative) division ring}.$$

The two roles of "point." Coordinatization does not define a point — it keeps the Hilbert primitive and merely assigns it a numerical name. Note the point wears two hats here: a point of the base line $\ell$ (call it a line-point, $P\in \ell$) is a field element (the carrier is literally "$F$ = the points of $\ell$", so $P$ is the number $\overrightarrow{OP}$), while a point of the plane $\Pi$ (a plane-point, $Q\in \Pi$) becomes a pair $(x,y)\in F^2$ under the map $\Phi$ of §1.2. So the geometry manufactures its own number system out of the points of one line, then re-describes every plane-point by two such numbers.

To say precisely which field object represents a given point, name the representation maps explicitly. It is cleanest to present the carrier abstractly as the set

$$F=\{[\overrightarrow{AB}]:A,B\in \ell \}/\cong$$

of congruence classes of directed segments of $\ell$ (with $[\overrightarrow{AA}]=0$ and $[\overrightarrow{OU}]=1$), so that "point" and "field element" are genuinely different kinds of object and the representation is a nontrivial map.

Definition (representation map on a line). The coordinate of a line-point $P\in \ell$ is $$\kappa :\ell ⟶F,\kappa (P)=[\overrightarrow{OP}],$$ the class of the directed segment from the origin to $P$. Then $\kappa (O)=0$, $\kappa (U)=1$, and $\kappa$ is a bijection (injective because $A\ne B>\Rightarrow \overrightarrow{OA}\ncong \overrightarrow{OB}$; surjective because every class has a representative with tail $O$). The field object representing $P$ is $\kappa (P)$.

The plane-point representation $\Phi :\Pi \to F^2$ is assembled from two copies of $\kappa$ (one per axis) in §1.2; we treat it there.

The map $\kappa$ is not just a bijection of sets but an isomorphism of ordered fields onto its image: it carries the geometric segment operations to the field operations,

$$\kappa (a+b)=\kappa (a)+\kappa (b),\kappa (a\cdot b)=\kappa (a)\kappa (b),a<b⟺\kappa (a)<\kappa (b),$$

which is precisely what the Hilbert theorem above asserts. (Under the concrete presentation $F:=\ell$ of the Construction, $\kappa ={\mathrm{id}}_{\ell }$ and a line-point simply is its own field object — the two presentations are isomorphic via $\kappa$, so nothing depends on the choice.) Thus "the field object represents a line-point" means exactly $P\mapsto \kappa (P)\in F$; the plane case $Q\mapsto \Phi (Q)\in F^2$ follows in §1.2.

The two named theorems are the coordinatization theorems: the geometry dictates the algebra. Verifying (O1)–(O2) and the group axioms is exactly the content of the segment-transport and similar-triangle lemmas.

1.2 The coordinate isomorphism

The field is built on one line; the whole plane is recovered by a bijection onto $F^2$ that turns geometric primitives into algebra.

The map. Fix two distinct lines through $O$ as axes ${\ell }_x,{\ell }_y$ (perpendicular, in Euclidean geometry), each carrying a copy of $F$ by §1.1 with the shared origin $O=0$ and a common unit point; write ${\kappa }_x,{\kappa }_y$ for their line-coordinates (§1.1). Given a point $P\in \Pi$, the parallel axiom provides a unique line through $P$ parallel to ${\ell }_y$; it meets ${\ell }_x$ in a single point ${\pi }_x(P)$ (incidence axioms), whose field-value is the abscissa $x(P)={\kappa }_x({\pi }_{x}P)\in F$. The ordinate $y(P)={\kappa }_y({\pi }_{y}P)$ is defined symmetrically using the parallel to ${\ell }_x$. Uniqueness of these parallels — hence well-definedness of $x(P),y(P)$ — is exactly Playfair's form of the parallel postulate; this is where the fifth postulate enters coordinatization. Define

$$\Phi :\Pi ⟶F^2,P⟼(x(P),y(P))=({\kappa }_x({\pi }_{x}P),{\kappa }_y({\pi }_{y}P)).$$

Bijectivity. $\Phi$ has an explicit inverse. Given $(a,b)\in F^2$, let ${\ell }_a$ be the line through the point $a\in {\ell }_x$ parallel to ${\ell }_y$, and ${\ell }_b$ the line through $b\in {\ell }_y$ parallel to ${\ell }_x$. Since ${\ell }_a\nparallel {\ell }_b$, they meet in a unique point $P$ (incidence + parallel axiom), and ${\Phi }^{-1}(a,b)=P$. Thus $\Phi$ is a bijection; the plane is set-theoretically $F^2$.

Structure preserved. $\Phi$ is not merely a bijection but an isomorphism of structures: it carries each geometric primitive to an algebraic condition on coordinates. Writing $P_i\mapsto (x_i,y_i)$,

The affine rows (lines, collinearity, betweenness, ratios along a line) require only the incidence, order, and parallel axioms — they are consequences of $\Phi$ respecting the linear structure of $F^2$. The metric row (equidistance) is where the congruence axioms III do their work: they force the quadratic form to be the displayed sum of squares once the axes are chosen perpendicular with a common unit. With oblique axes or unequal units one still gets an isomorphic geometry, but the form becomes a general positive-definite quadratic $q(x,y)=a{x}^2+2bxy+c{y}^2$; choosing an orthonormal frame is precisely diagonalizing $q$ to $x^2+y^2$.

What "isomorphism" means here. Let $(F^2,B^{\ast },{\equiv }^{\ast })$ denote the set $F^2$ equipped with the algebraically defined betweenness and equidistance of the table. Then:

Representation theorem. $\Phi :(\Pi ,B,\equiv )\to (F^2,B^{\ast },>{\equiv }^{\ast })$ is an isomorphism of relational structures. Consequently every model of the (first-order) plane-geometry axioms with coordinate field $F$ is isomorphic to the standard analytic plane $F^2$ over that field.

This is the exact sense in which "the plane maps to the field": the points of $\ell$ are in bijection with the elements of $F$, and the points of $\Pi$ are in bijection with $F^2$, compatibly with all the geometric relations. It is one half of the bi-interpretation of §3 — the geometry is interpreted in the ordered field $(F,+,\cdot ,<)$, and (§3) conversely.

Coordinate independence. The construction depends on choices — the origin $O$, the axes, the unit — but the geometry does not. Two coordinatizations $\Phi ,{\Phi }^{\prime }$ differ by a composite affine map ${\Phi }^{\prime }\circ {\Phi }^{-1}:F^2\to F^2$, $\mathbf{x}\mapsto A\mathbf{x}+\mathbf{b}$ with $A\in G{L}_2(F)$; restricting to frames that preserve the equidistance form gives $A\in O_2(F)$ and the Euclidean isometry group $E(2)=F^2⋊O_2(F)$. So coordinates are a gauge: everything geometric is invariant under the change-of-frame group, exactly as in the Erlangen picture. The same construction in $n$ dimensions yields a bijection ${\Pi }_n\cong F^n$.

Representing lines and planes. A point is represented by a field tuple; the other Hilbert primitives get field representatives too — but of a dual kind, namely the coefficients of their defining equation, taken up to scaling.

• Lines. Let $\mathcal{L}$ be the set of lines of $\Pi$. By the table, $\Phi$ sends a line $g$ to a solution set $$\Phi (g)=\{(x,y)\in F^2:\alpha x+\beta y=\gamma \},(\alpha ,\beta )\ne (0,0).$$ The triple $(\alpha ,\beta ,\gamma )$ is determined by $g$ only up to a common nonzero scalar $\lambda \in F^{\times }$ (scaling the equation gives the same set), so the representation of lines is $$\lambda :\mathcal{L}⟶(\{(\alpha ,\beta ,\gamma )\in F^3:(\alpha ,\beta )\ne 0\})/F^{\times },g⟼[\alpha :\beta :\gamma ],$$ a bijection onto these classes — a subset of the projective plane ${\mathbb{P}}^2(F)$ (all of it except the class $[0:0:1]$). The field object representing a line is thus a projective coordinate triple $[\alpha :\beta :\gamma ]$, dual to the affine point-coordinates.
• Incidence and duality. Homogenizing a point as $[x:y:1]$, incidence $P\in g$ becomes the symmetric vanishing pairing $$\alpha x+\beta y-\gamma \cdot 1=0,$$ perfectly symmetric between the point triple $(x,y,1)$ and the line triple $(\alpha ,\beta ,-\gamma )$. This is the algebraic source of projective duality: points and lines are both classes of triples, interchanged by swapping the two roles.
• Planes. In the plane there is only one plane, $\Pi$ itself, represented by all of $F^2$; the primitive "plane" earns a nontrivial representative in solid geometry. There points are triples, ${\Phi }_3:{\Pi }_3\stackrel{\sim }{\to }{F}^3$, and a plane is a solution set $$\{(x,y,z)\in F^3:\alpha x+\beta y+\gamma z=\delta \},(\alpha ,\beta ,\gamma )\ne 0,$$ represented up to scaling by $[\alpha :\beta :\gamma :\delta ]\in {\mathbb{P}}^3(F)$. In general, in $n$ dimensions a hyperplane is represented by a class $[{\alpha }_0:\cdots :{\alpha }_n]$ dual to the homogeneous point coordinates $[x_1:\cdots :x_n:1]$, with incidence the vanishing of the symmetric pairing ${\alpha }_0{x}_1+\cdots$. Lines in space are not hyperplanes and need a separate (Plücker/Grassmann) representation — beyond this page.

So the full dictionary of primitives is: a point $\mapsto$ a tuple in $F^n$ (§1.1–1.2), while a line or plane $\mapsto$ a projective class of equation coefficients — points and hyperplanes being dual species of the same field data.

1.3 Which field?

How rich $F$ is depends on how strong the axioms are:

Via $\Phi$ the plane is $F^2$, and Euclidean geometry becomes coordinate geometry over $F$, with distance $\mathrm{dist}(P,Q{)}^2=(x_1-x_2{)}^2+(y_1-y_2{)}^2$ — the analytic form of Pythagoras that the metric-geometry page takes as its flat metric.

Worked example: the standard plane $F=\mathbb{R}$. Take $\Pi ={\mathbb{R}}^2$, base line $\ell$ = the $x$-axis, $O=(0,0)$, $U=(1,0)$. Every ingredient reduces to the familiar real operations:

• Carrier & $\kappa$. The class of $\overrightarrow{OP}$ is the signed length, so $F=\mathbb{R}$, and the line coordinate reads it off: $$\kappa :\{(t,0)\}\to \mathbb{R},\kappa ((t,0))=t,\kappa (O)=0,\kappa (U)=1.$$ Under the concrete presentation $F:=\ell$ this $\kappa$ is the identity $\mathbb{R}\to \mathbb{R}$; the $y$-axis carries the symmetric copy ${\kappa }_y((0,s))=s$.
• Addition is transport along $\ell$: $(a,0)\mapsto (a+b,0)$, i.e. ordinary real $+$; the inverse is the reflection $(-a,0)$.
• Order is betweenness on the axis — the usual $<$ on $\mathbb{R}$.
• Multiplication is the intercept theorem. With auxiliary line $m$ = the $y$-axis, put $E=(0,1)$, $A=(0,a)$ on $m$ and $B=(b,0)$ on $\ell$; the parallel to $EB$ through $A$ meets $\ell$ at $C=(ab,0)$, since $OC:OB=OA:OE$ gives $OC=ab$. E.g. $a=2,b=3$: line $EB$ is $\frac{x}{3}+y=1$, its parallel through $A=(0,2)$ is $\frac{x}{3}+y=2$, meeting $y=0$ at $x=6$ — the construction returns $2\cdot 3=6$.
• Coordinate map $\Phi$. For $Q=(x,y)$, the parallel-projections are ${\pi }_x(Q)=(x,0)$ (meet of the $x$-axis with the vertical through $Q$) and ${\pi }_y(Q)=(0,y)$, so $$\Phi (Q)=({\kappa }_x({\pi }_{x}Q),{\kappa }_y({\pi }_{y}Q))=(x,y).$$ With the standard axes $\Phi :{\mathbb{R}}^2\to {\mathbb{R}}^2$ is thus the identity (ordinary Cartesian coordinates); e.g. $Q=(2,3)\mapsto (2,3)$. The distance is the usual $\sqrt{(x_1-x_2{)}^2+(y_1-y_2{)}^2}$. That $\kappa$ and $\Phi$ come out as identities is an artifact of choosing the standard axes and the concrete presentation $F:=\ell$: a rotated/translated frame, or the abstract segment-class carrier, makes both genuine (nontrivial) bijections, with $\Phi$ an affine map $\mathbf{x}\mapsto A\mathbf{x}+\mathbf{b}$ per §1.2.

The primitives, spelled out in this model:

• Point $=$ a pair $(x,y)\in {\mathbb{R}}^2$ (in solid geometry, a triple in ${\mathbb{R}}^3$). E.g. $(2,3)$.
• Line $=$ a solution set $\{(x,y):\alpha x+\beta y=\gamma \}$, $(\alpha ,\beta )\ne (0,0)$ — the usual lines $y=mx+c$ and verticals $x=c$ — represented by the projective class $[\alpha :\beta :\gamma ]\in {\mathbb{RP}}^2>\setminus \{[0:0:1]\}$. E.g. $x+y=1$ is $[1:1:1]$, the $x$-axis is $[0:1:0]$, the vertical $x=5$ is $[1:0:-5]$.
• Plane $=$ all of ${\mathbb{R}}^2$ — there is only one. A nontrivial plane-object appears one dimension up: in ${\mathbb{R}}^3$ a plane $\{\alpha x+\beta y+\gamma z=\delta \}$ is $[\alpha :\beta :\gamma :\delta ]\in >{\mathbb{RP}}^3$; e.g. $z=0$ is $[0:0:1:0]$.
• Incidence. $(x,y)$ lies on $[\alpha :\beta :\gamma ]$ iff $\alpha x+\beta y->\gamma =0$; e.g. $(2,3)\in [1:1:5]$ since $2+3-5=0$.

$\mathbb{R}$ is Dedekind-complete, so it sits in the last row of the table (and is in particular real closed). The point of the construction is that these real operations are recovered synthetically — from betweenness and congruence — with no numbers presupposed.

2. Real Closed Fields

An ordered field $F$ is real closed if it is first-order indistinguishable from $\mathbb{R}$. Equivalent characterizations:

• every positive element has a square root and every odd-degree polynomial has a root; equivalently
• $F$ is not algebraically closed but $F[\sqrt{-1}]$ is (so $F$ sits "one step below" algebraically closed); equivalently
• $F\equiv \mathbb{R}$ (same first-order theory) — this theory is RCF.

Crucially there are many real closed fields, all modelling RCF and so all carrying "the same" elementary Euclidean geometry:

• $\mathbb{R}$ itself;
• the real algebraic numbers ${\mathbb{R}}_{\mathrm{alg}}=\mathbb{R}\cap \overline{\mathbb{Q}}$ — countable and computable;
• non-Archimedean fields: the hyperreals, Puiseux series, Hahn series, containing genuine infinitesimal segments.

That RCF has models other than $\mathbb{R}$ is the Löwenheim–Skolem phenomenon, and it is why pinning down $\mathbb{R}$ uniquely requires the second-order continuity axiom (last row of §1's table).

3. The Bi-Interpretation

The precise sense in which "geometry $=$ theory of $\mathbb{R}$" is bi-interpretability: each theory is definable inside the other, and the two translations are mutually inverse. Tarski's geometry uses one sort (points) and two primitives — betweenness $B(a,b,c)$ and equidistance $ab\equiv cd$ (the first-order axiomatization).

What is a point here? The only primitive. Unlike Hilbert's three sorts (points, lines, planes), Tarski's first-order geometry has a single sort: points. There is no primitive "line": a point is the sole primitive — in the standard model an element of $F^2$, i.e. a pair of field elements — and a line is a definable set of points $\{(x,y):\alpha x+\beta y=\gamma \}$, as are circles, segments, and congruence. So "inside the theory of $\mathbb{R}$" a point is an element of $F^2$ and everything else is defined from points by polynomial conditions. This single-sort economy is what makes the interpretation below so clean: to interpret geometry in the field you need only say what a point is ($F^2$) and how the point-relations translate — there is no separate line object to account for. (The pair $(x,y)$ is a point's address relative to a chosen frame, not its intrinsic identity: changing the frame acts by an affine map, per §1.2.)

Geometry $\to$ field. Fix an origin, a unit, and perpendicular axes (all definable from $B$ and $\equiv$). Coordinates of points along an axis carry definable operations: segment addition gives $+$, the similar-triangles product gives $\cdot$, betweenness gives $<$. This interprets $(\mathbb{R},+,\cdot ,<)$ inside the geometry (§1 made this constructive).

Field $\to$ geometry. Conversely, set points $=F^2$ and define the primitives algebraically:

$$ab\equiv cd:⟺(a_1-b_1{)}^2+(a_2-b_2{)}^2=(c_1-d_1{)}^2+(c_2-d_2{)}^2,$$

with betweenness the analogous polynomial condition. This interprets the geometry inside RCF.

Because the two interpretations are inverse (up to definable isomorphism), the theories share every model-theoretic invariant: completeness, decidability, quantifier elimination, o-minimality, the same definable sets, the same models. This is the exact content of the slogan.

4. Quantifier Elimination = Semialgebraic Geometry

Tarski's theorem (1930s): RCF admits quantifier elimination in the ordered-field language, and is therefore complete and decidable. Its geometric meaning is the Tarski–Seidenberg theorem. Call a set semialgebraic if it is a Boolean combination of polynomial equalities and inequalities. Then:

Tarski–Seidenberg. The projection of a semialgebraic set is semialgebraic.

Eliminating an existential quantifier $\exists x_n$ is forgetting a coordinate, i.e. projecting a figure onto the remaining axes. Consequences:

• the definable sets of the geometry are exactly the semialgebraic sets — lines, circles, conics, polygons, and everything built from them by finite unions, intersections, complements, and projections;
• every first-order geometric statement is equivalent to a quantifier-free polynomial condition on the coordinates.

RCF is thereby the flagship example of o-minimality (see model-theory.md): definable subsets of a line are finite unions of points and intervals — the exact formalization of the classical intuition that a line meets a curve in finitely many points or lies along a whole segment. O-minimality is the modern model-theoretic theory of tame geometry.

5. Decidability and Its Limits

• A decision procedure for Euclid. Because RCF is decidable, there is an algorithm (Tarski's procedure; in practice Collins' cylindrical algebraic decomposition) deciding the truth of any sentence of elementary geometry — "do the three perpendicular bisectors of a triangle always concur?" is mechanically checkable. Elementary geometry is, in principle, automatable.
• A canonical smallest model. ${\mathbb{R}}_{\mathrm{alg}}$ is the prime model of RCF: it elementarily embeds into every model. Every straightedge-and-compass point already lives there, which is why the decision procedure needs no genuine real analysis.
• Not categorical. As in §2, non-Archimedean real closed fields give geometries with infinitesimal segments; first-order geometry cannot see the difference. Uniqueness of the Euclidean plane needs the second-order continuity axiom.
• No conflict with Gödel. Elementary geometry is complete and decidable because it is too weak to define $\mathbb{N}$: there is no first-order way to say "$x$ is an integer" from $B$ and $\equiv$ alone, so it cannot encode arithmetic. Adjoin a predicate for $\mathbb{Z}$ and $(\mathbb{R},+,\cdot ,\mathbb{Z})$ becomes undecidable. The incompleteness theorems bite on arithmetic, not on geometry — a striking inversion of the usual worry.

6. One Field, All Three Geometries

Because everything is definable over the field, the coordinate picture is not special to the flat case. Over any real closed field $F$:

• Euclidean geometry is $F^2$ with the metric of §1;
• the Cayley–Klein construction — fix an absolute conic, measure by cross-ratio — realizes hyperbolic and elliptic geometry as definable structures over the same $F$.

So "the theory of $\mathbb{R}$" is a single model-theoretic home for the entire geometry spine, and the independence of the parallel postulate becomes a statement about which conic you interpret, not about which field you use. The next page keeps the coordinates but replaces the global polynomial distance with a local metric tensor, opening the door to variable curvature and the Riemannian synthesis.

7. Summary

Coordinatization turns Euclid's synthetic axioms into the algebra of a real closed field; Tarski's theorem turns that algebra into a decision procedure; and the whole of elementary geometry — Euclidean and non-Euclidean alike — becomes the study of semialgebraic sets over $\mathbb{R}$.