Ordered Fields

The ordered-field axioms are the silent scaffolding underneath every inequality manipulation in calculus, from solving $|x-3|<5$ to justifying the squeeze theorem. In real analysis, the compatibility rules proved here are exactly what license algebraic manipulation of $\epsilon$-inequalities in limit proofs, and they are the base layer on which absolute value and later the completeness axiom are built to get from $\mathbb{Q}$ to $\mathbb{R}$. In algebra, the Artin–Schreier theory of formally real fields (which of the counterexample above is a first instance) governs which fields can carry a compatible order at all — a question central to real algebraic geometry, where one studies solution sets of polynomial inequalities over ordered fields. In logic and model theory, "real closed fields" (ordered fields satisfying an additional first-order completeness-like scheme) underlie quantifier elimination algorithms used in computer algebra systems and automated theorem proving. In optimization and control theory, linear and semidefinite programming over the reals depend at every step on the sign-preservation rules of Theorem parts (1)–(3) to justify manipulating constraint inequalities. In cryptography, the impossibility of ordering finite fields (illustrated by the $\mathbb{Z}/3\mathbb{Z}$ remark in the counterexample) is one reason finite-field arithmetic (used in elliptic-curve cryptography) behaves so differently from real-number arithmetic — there is no notion of "positive" or "less than" to exploit or defend against.

Ask the tutor: How does the completeness axiom get added on top of the ordered-field axioms to single out $\mathbb{R}$ among all ordered fields?

Think of a field as a machine with two operations obeying algebraic laws (associativity, distributivity, inverses), entirely blind to any notion of "bigger" or "smaller." An order relation is, on its own, just a way of lining up elements on an abstract line, with no reference to $+$ or $\cdot$ at all — trichotomy and transitivity alone would let you order the field arbitrarily, in a way with no connection to the arithmetic. The content of an ordered field is that the lining-up is forced to respect the arithmetic. Axiom (O3) says the order is translation invariant: sliding the whole number line by $z$ doesn't change which of two points is to the right. Axiom (O4) says multiplying by a positive number is a order-preserving rescaling: it can stretch or shrink distances, but a point to the right of another stays to the right, as long as you rescale by something positive. These two axioms alone force an enormous amount of structure — including the fact, proved below, that $-1$ must be negative, and hence that no square root of $-1$ can exist in an ordered field. The picture to keep in mind is the ordinary number line for $\mathbb{Q}$: shifting every point right by $z$ preserves left-right order (O3), and stretching the line by a positive factor $z$ preserves left-right order too (O4), while stretching by a negative factor flips the whole line over — a fact we will derive, not assume.

Example 1 (checking compatibility numerically in $\mathbb{Q}$). Let $F=\mathbb{Q}$ with its usual order. Take $a=-\tfrac{5}{3}$, $b=\tfrac{2}{7}$, $c=-4$. We verify Theorem parts (1) and (3) with these asymmetric numbers. Part (1) check. We need a second pair; take $c'=\tfrac{1}{9}$, $d'=\tfrac{1}{2}$ (indeed $c'<d'$ since $\tfrac1{9}=0.111\ldots < 0.5$). Since $a<b$ (as $-1.\overline{6}<0.2857\ldots$), part (1) predicts

Compute: $a+c' = -\tfrac{5}{3}+\tfrac19 = -\tfrac{15}{9}+\tfrac19=-\tfrac{14}{9}\approx -1.5556$. And $b+d'=\tfrac27+\tfrac12=\tfrac{4}{14}+\tfrac{7}{14}=\tfrac{11}{14}\approx 0.7857$. Indeed $-1.5556<0.7857$. ✓. Part (3) check. With $a<b$ as above and $c=-4<0$, the theorem predicts $bc<ac$, i.e. multiplying by the negative number $-4$ reverses the order. Compute $ac = \left(-\tfrac53\right)(-4)=\tfrac{20}{3}\approx 6.667$, and $bc=\left(\tfrac27\right)(-4)=-\tfrac{8}{7}\approx -1.143$. Indeed $bc=-1.143 < ac=6.667$, confirming the reversal, and matching the direction "the originally smaller number ($a$) now yields the larger product ($ac$)." ✓

Example 2 (solving a rational inequality using only the axioms/theorem, not calculus). Solve, over $\mathbb{Q}$, the inequality

for $x\neq -3$, citing only field-and-order facts (parts of the Theorem) at each step — no appeal to continuity or graphing. Step 1. Rewrite: $\dfrac{2x-7}{x+3}-1<0 \iff \dfrac{2x-7-(x+3)}{x+3}<0 \iff \dfrac{x-10}{x+3}<0$, using only field axioms (common denominator, distributivity) — this is an equality of expressions, not yet an order step. Step 2. A quotient $\dfrac{x-10}{x+3}$ is negative exactly when numerator and denominator have opposite (strict) signs. This is itself a consequence of the Theorem: if $u,v$ have opposite signs, say $u>0,v<0$, then $v^{-1}<0$ (by the same sign-of-inverse argument used in part (5), noting $v^{-1}\cdot v=1>0$ forces $v^{-1}$ and $v$ to share a sign, since a negative times a positive is negative by part (3), yet the product must be the positive number $1$ — so $v^{-1}<0$ when $v<0$). Then $u\cdot v^{-1}$ is a product of a positive and a negative, hence negative by part (3) (taking $c=v^{-1}<0$, $a=0<u$, giving $u\cdot v^{-1}<0\cdot v^{-1}=0$). Symmetrically opposite signs the other way also give a negative quotient, and like signs give a positive quotient (product of two positives, or two negatives, is positive — the latter since $(-p)(-q)=pq$ for $p,q>0$ by field-axiom consequences, and $pq>0$ by (O4)). Step 3. Apply Step 2 to $u=x-10$, $v=x+3$. Two cases give a negative quotient:

• Case A: $x-10>0$ and $x+3<0$. The first gives $x>10$; the second gives $x<-3$. No $x$ satisfies both (if $x>10$ then certainly not $x<-3$, by transitivity (O2): $x<-3<10<x$ would give $x<x$, impossible by (O1)). Case A is empty.
• Case B: $x-10<0$ and $x+3>0$. The first gives $x<10$; the second gives $x>-3$. Both hold exactly for $-3<x<10$.

Conclusion. The solution set is $\{x\in\mathbb{Q} : -3<x<10\}$. As a numerical check take $x=4$ (asymmetric, inside the interval): $\frac{2(4)-7}{4+3}=\frac{1}{7}\approx 0.143<1$. ✓. Take $x=20$ (outside): $\frac{2(20)-7}{20+3}=\frac{33}{23}\approx 1.435 \not< 1$. ✓ consistent with exclusion.