Consequences of the Order Axioms

Why is a negative number times a negative number positive? Every student learns this rule. But almost no one is shown a proof. It turns out to be a short, clean deduction from a small list of axioms. The exercise of proving it is the first real test. It shows whether you understand what an ordered field actually is. The real numbers are built to satisfy a handful of order axioms. Everything else is a theorem, not an assumption. This includes the sign rules for arithmetic, the fact that squares are never negative, and the behavior of inequalities under multiplication and division. This matters because so much of analysis is inequality-chasing. This includes bounding errors, comparing sequences, and squeezing limits. None of that is trustworthy unless the underlying rules of inequality manipulation are airtight. Suppose you don't know why multiplying both sides of an inequality by a negative number reverses it. Then you are applying a rule you can't defend. Eventually you will misapply it. The big picture is this: an ordered field is a field FF. It comes together with a subset P⊂FP \subset F of "positive" elements. This subset satisfies just three closure/trichotomy axioms. From these three axioms alone, one can derive the entire arithmetic of inequalities. This requires no reference to decimals. It needs no geometric picture of a number line. It needs no appeal to "obviously true." The real number line is one model of these axioms. But the theorems below hold in any ordered field, including the rational numbers. This article delivers rigorous proofs of the standard sign rules. These include: negative times negative is positive, and positive times negative is negative. It also proves that every nonzero square is strictly positive. Hence 1>01 > 0 in any ordered field. Finally, it covers the core algebraic rules for manipulating inequalities. These rules apply under addition, multiplication, and taking reciprocals. Each proof is built only from the order axioms and basic field axioms. There are no circular appeals to facts we have not yet established.

Key takeaways
  • An ordered field has a positivity set PP closed under +,×+,\times, with exactly one of a∈Pa\in P, a=0a=0, −a∈P-a\in P holding for each aa (trichotomy).
  • From this alone: a<b  ⟺  b−a∈Pa<b \iff b-a\in P defines a genuine order relation, transitive and compatible with addition.
  • Sign rules follow rigorously: (−a)(−b)=ab(-a)(-b)=ab, and (−a)b=−(ab)(-a)b=-(ab); in particular negative ×\times negative is positive.
  • Every nonzero element has a strictly positive square, so 1>01>0 always, and −1-1 is never a square.
  • Multiplying an inequality by a positive element preserves it; multiplying by a negative element reverses it.
  • Key idea: every familiar sign/inequality rule is a finite deduction from trichotomy and closure of PP under sums and products — nothing more is assumed.

Where this lives today

The Artin–Schreier theory of formally real fields, foreshadowed by the counterexample above, is a cornerstone of real algebraic geometry and model theory, used to classify which fields can carry an order at all. In numerical analysis and interval arithmetic, the sign rules proved here are exactly the logical justification for how software correctly propagates error bounds through multiplications and divisions. In cryptography, ordered-field reasoning underlies correctness proofs for algorithms over ordered rings (e.g., lattice-based schemes working with bounded integers, where sign and magnitude tracking must be airtight). In physics and engineering, dimensional and inequality reasoning (stability margins, error bounds in control theory) implicitly relies on exactly the monotonicity laws (O7)–(O8) established here. Pedagogically, this material is the direct bridge to Absolute Value and Its Properties, where ∣a∣|a| is defined using trichotomy, and to Bounds of a Set, where suprema and infima are defined and compared using the very inequality relation << constructed in this article.

Ask the tutor: Why does the reciprocal-reversal rule (O8) fail if aa and bb don't have the same sign?

DefinitionOrdered Field

A field FF is an ordered field if there exists a subset P⊂FP \subset F (the positive elements) such that: 1. (Closure under addition) If a,b∈Pa, b \in P, then a+b∈Pa + b \in P. 2. (Closure under multiplication) If a,b∈Pa, b \in P, then ab∈Pab \in P. 3. (Trichotomy) For every a∈Fa \in F, exactly one of the following holds:

a∈P,a=0,−a∈P.a \in P, \qquad a = 0, \qquad -a \in P.

We define the relation << on FF by

a<biffb−a∈P,a < b \quad \text{iff} \quad b - a \in P,

and write a>ba > b for b<ab < a, a≤ba \le b for (a<ba < b or a=ba = b), and a≥ba \ge b similarly. We also write a>0a > 0 to mean a∈Pa \in P, matching the usual notion of "positive."

Intuition

Think of PP as a rule that colors certain field elements "positive" without yet knowing what a number line looks like. The three axioms are exactly the minimal demands needed to make that coloring behave the way we want: positives should stay positive under addition and multiplication (closure), and every element should be positive, zero, or the negative of something positive, with no overlaps (trichotomy). Everything downstream — "less than," "greater than," the sign rules — is manufactured from this coloring rather than assumed directly. The picture to hold in your head is a number line split into three disjoint pieces: the positive ray, the single point 00, and the negative ray, which is the mirror image of the positive ray under negation. The closure axioms say the positive ray is closed under addition and multiplication — add or multiply two positive things and you stay in the positive ray. The proofs below are simply careful bookkeeping: tracking which of the three regions a given expression like (−a)(−b)(-a)(-b) or a2a^2 falls into, using only algebra and the axioms.

Figure
The trichotomy partition of a field F into negatives, zero, and positives
0 P (positives) −P (negatives) −a ∈ P a ∈ P a ↦ −a mirrors the ray
TheoremBasic Order Properties of an Ordered Field

Let FF be an ordered field with positive set PP, and let a,b,c∈Fa, b, c \in F. Then:

(O1) Irreflexivity/Trichotomy for <<. For any a,b∈Fa,b \in F, exactly one of a<ba<b, a=ba=b, a>ba>b holds.

(O2) Transitivity. If a<ba < b and b<cb < c, then a<ca < c.

(O3) Translation invariance. If a<ba < b, then a+c<b+ca + c < b + c for every c∈Fc \in F.

(O4) Sign of negation. a>0a > 0 if and only if −a<0-a < 0.

(O5) Sign rules for products. (a) If a>0a > 0 and b>0b > 0, then ab>0ab > 0. (b) If a>0a > 0 and b<0b < 0, then ab<0ab < 0. (c) If a<0a < 0 and b<0b < 0, then ab>0ab > 0. In particular, (−a)(−b)=ab(-a)(-b) = ab for all a,b∈Fa,b \in F, and (−1)a=−a(-1)a = -a.

(O6) Squares are non-negative. For every a∈Fa \in F, a2≥0a^2 \ge 0, with equality iff a=0a = 0. Consequently 1>01 > 0.

(O7) Multiplicative monotonicity. If a<ba < b and c>0c > 0, then ac<bcac < bc. If a<ba < b and c<0c < 0, then ac>bcac > bc.

(O8) Reciprocals reverse order for same-sign elements. If 0<a<b0 < a < b, then 0<1b<1a0 < \frac{1}{b} < \frac{1}{a}.

Proof

We work only from the field axioms of FF and the three defining properties of PP (closure under ++, closure under ×\times, trichotomy), plus the definition a<b  ⟺  b−a∈Pa<b \iff b-a \in P. Throughout, we use the standard field facts (−1)a=−a(-1)a = -a, −(−a)=a-(-a)=a, and −(ab)=(−a)b=a(−b)-(ab) = (-a)b = a(-b); these are pure field-axiom consequences (proved by the usual distributive-law manipulations, independent of order) and we take them as already established, since they involve no order concept.

Proof of (O1). By the trichotomy axiom applied to the element b−a∈Fb - a \in F, exactly one of

b−a∈P,b−a=0,−(b−a)=a−b∈Pb - a \in P, \qquad b-a = 0, \qquad -(b-a) = a - b \in P

holds. These are precisely the statements a<ba < b, a=ba = b, b<ab < a (i.e. a>ba>b) respectively. Hence exactly one of a<ba<b, a=ba=b, a>ba>b holds. ■\blacksquare (O1)

Proof of (O2). Suppose a<ba<b and b<cb<c. By definition, b−a∈Pb - a \in P and c−b∈Pc - b \in P. By closure of PP under addition,

(b−a)+(c−b)∈P.(b-a) + (c-b) \in P.

Field algebra gives (b−a)+(c−b)=c−a(b-a)+(c-b) = c - a. Hence c−a∈Pc - a \in P, i.e., a<ca < c. ■\blacksquare (O2)

Proof of (O3). Suppose a<ba < b, so b−a∈Pb - a \in P. For any c∈Fc \in F,

(b+c)−(a+c)=b−a∈P,(b+c) - (a+c) = b - a \in P,

by field algebra (the cc's cancel). By definition this says a+c<b+ca + c < b + c. ■\blacksquare (O3)

Proof of (O4). By definition, a>0a > 0 means a−0=a∈Pa - 0 = a \in P. Also −a<0-a < 0 means 0−(−a)=a∈P0 - (-a) = a \in P. These are literally the same statement, a∈Pa \in P, so they are equivalent. ■\blacksquare (O4)

Proof of (O5). (a) Suppose a>0a > 0 and b>0b > 0, i.e., a∈Pa \in P and b∈Pb \in P. By closure of PP under multiplication, ab∈Pab \in P, i.e., ab>0ab > 0. (b) Suppose a>0a > 0 and b<0b < 0. Then a∈Pa \in P, and by (O4), b<0b < 0 means −b∈P-b \in P. By closure under multiplication, a(−b)∈Pa(-b) \in P. But by field algebra a(−b)=−(ab)a(-b) = -(ab), so −(ab)∈P-(ab) \in P, which by definition means ab<0ab < 0. (c) Suppose a<0a < 0 and b<0b < 0. By (O4), −a∈P-a \in P and −b∈P-b \in P. By closure under multiplication, (−a)(−b)∈P(-a)(-b) \in P. By field algebra,

(−a)(−b)=−(a(−b))=−(−(ab))=ab,(-a)(-b) = -(a(-b)) = -(-(ab)) = ab,

using −(ab)=a(−b)-(ab)=a(-b) (field algebra) twice and the involution −(−x)=x-(-x)=x. Hence ab∈Pab \in P, i.e., ab>0ab > 0. For the "in particular" clause: the identity (−a)(−b)=ab(-a)(-b) = ab was derived above purely by field algebra together with part (c)'s computation (it holds for all a,b∈Fa, b \in F, not just negative ones, since it's a pure algebraic identity — we only needed the sign hypothesis to conclude positivity, not to derive the identity itself). Setting b=1b=1: (−a)(−1)=a⋅1=a(-a)(-1) = a\cdot 1 = a; combined with the general identity (−a)(−1)=−(−a)⋅(-a)(-1) = -(-a) \cdot [algebra] one recovers (−1)a=−a(-1)a = -a directly from field axioms as noted above. ■\blacksquare (O5)

Proof of (O6). Let a∈Fa \in F. By trichotomy exactly one of a∈Pa \in P, a=0a = 0, −a∈P-a \in P holds.

  • If a∈Pa \in P (i.e. a>0a > 0): by (O5)(a) with b=ab = a, a⋅a=a2∈Pa \cdot a = a^2 \in P, so a2>0a^2 > 0.
  • If a=0a = 0: then a2=0⋅0=0a^2 = 0 \cdot 0 = 0 by field algebra.
  • If −a∈P-a \in P (i.e. a<0a < 0): by (O5)(c) with b=ab=a, a⋅a=a2∈Pa \cdot a = a^2 \in P, so a2>0a^2 > 0.

In every case a2≥0a^2 \ge 0, and the only case giving a2=0a^2 = 0 is a=0a = 0; conversely if a=0a=0 then a2=0a^2=0. This proves the equivalence. For the consequence 1>01 > 0: since FF is a field, 1≠01 \neq 0 (field axiom: 0≠10 \neq 1). Apply the above with a=1a = 1: since 1≠01 \ne 0, 12=1⋅1=1∈P1^2 = 1 \cdot 1 = 1 \in P by the dichotomy just shown (as 12>01^2 > 0 whenever 1≠01 \ne 0). Hence 1>01 > 0. ■\blacksquare (O6)

Proof of (O7). Suppose a<ba < b, so b−a∈Pb - a \in P. First claim: if c>0c > 0, then ac<bcac < bc. Since c∈Pc \in P and b−a∈Pb - a \in P, closure under multiplication gives (b−a)c∈P(b-a)c \in P. By the distributive law, (b−a)c=bc−ac(b-a)c = bc - ac. Hence bc−ac∈Pbc - ac \in P, i.e. ac<bcac < bc. Second claim: if c<0c < 0, then ac>bcac > bc. By (O4), −c∈P-c \in P. Since b−a∈Pb - a \in P, closure under multiplication gives (b−a)(−c)∈P(b-a)(-c) \in P. By field algebra, (b−a)(−c)=−(b−a)c=−(bc−ac)=ac−bc(b-a)(-c) = -(b-a)c = -(bc - ac) = ac - bc. Hence ac−bc∈Pac - bc \in P, i.e., bc<acbc < ac, i.e., ac>bcac > bc. ■\blacksquare (O7)

Proof of (O8). Suppose 0<a<b0 < a < b. First, we show a−1,b−1a^{-1}, b^{-1} exist and are positive. Since a≠0a \neq 0 (as a∈Pa \in P and 0∉P0 \notin P by trichotomy applied to 00: if 0∈P0\in P then also −0=0∈P-0=0\in P, contradicting exactly-one-of-three unless we note 00 satisfies the middle clause, so 0∉P0 \notin P), aa has a multiplicative inverse a−1a^{-1} in the field FF. We claim a−1>0a^{-1} > 0. By trichotomy applied to a−1a^{-1}, either a−1∈Pa^{-1} \in P, or a−1=0a^{-1}=0, or −a−1∈P-a^{-1} \in P.

  • a−1=0a^{-1} = 0 is impossible: it would give a⋅a−1=a⋅0=0a \cdot a^{-1} = a \cdot 0 = 0, contradicting a⋅a−1=1≠0a\cdot a^{-1}=1\ne 0.
  • If −a−1∈P-a^{-1} \in P: since also a∈Pa \in P, closure under multiplication gives a(−a−1)∈Pa(-a^{-1}) \in P. But a(−a−1)=−(a⋅a−1)=−1a(-a^{-1}) = -(a\cdot a^{-1}) = -1. So −1∈P-1 \in P. But also 1∈P1 \in P by (O6), so by closure under addition 1+(−1)=0∈P1 + (-1) = 0 \in P. This contradicts trichotomy at 00 (we showed 0∉P0 \notin P above). Contradiction, so this case is impossible.

Hence by elimination a−1∈Pa^{-1} \in P, i.e. a−1>0a^{-1} > 0. The identical argument (verbatim, replacing aa by bb) gives b−1>0b^{-1} > 0. Now from a<ba < b, multiply both sides by c:=a−1b−1c := a^{-1}b^{-1}. Since a−1>0a^{-1} > 0 and b−1>0b^{-1} > 0, closure under multiplication gives c=a−1b−1∈Pc = a^{-1}b^{-1} \in P, i.e., c>0c > 0. By (O7) (first claim) applied with this positive cc,

a⋅(a−1b−1)<b⋅(a−1b−1).a \cdot (a^{-1}b^{-1}) < b \cdot (a^{-1}b^{-1}).

Simplify each side by field algebra: a⋅a−1b−1=(aa−1)b−1=1⋅b−1=b−1a\cdot a^{-1}b^{-1} = (a a^{-1})b^{-1} = 1\cdot b^{-1} = b^{-1}, and b⋅a−1b−1=a−1(bb−1)=a−1⋅1=a−1b \cdot a^{-1}b^{-1} = a^{-1}(b b^{-1}) = a^{-1}\cdot 1 = a^{-1}. So the inequality reads

b−1<a−1,b^{-1} < a^{-1},

i.e. 1b<1a\frac{1}{b} < \frac{1}{a}. Combined with b−1>0b^{-1}>0 shown above, we get 0<1b<1a0 < \frac1b < \frac1a. ■\blacksquare (O8)

∎

Worked examples

Example 1 — chaining sign rules on an asymmetric algebraic expression. Let FF be an ordered field and suppose x,y,z∈Fx, y, z \in F satisfy

x<0,0<y<z.x < 0, \qquad 0 < y < z.

We determine the sign of

E=x(z−y)+y2.E = x(z - y) + y^2.

Step 1: sign of z−yz - y. Since y<zy < z, by definition z−y∈Pz - y \in P, so z−y>0z - y > 0. Step 2: sign of x(z−y)x(z-y). We have x<0x < 0 and z−y>0z - y > 0. By (O5)(b) (with the roles of the two factors swapped, which is valid since multiplication in a field is commutative and the sign rule is symmetric in its statement — explicitly, (O5)(b) as proved covers "positive times negative"; here we apply it to the product (z−y)⋅x(z-y)\cdot x, positive times negative, giving (z−y)x<0(z-y)x < 0, and then x(z−y)=(z−y)xx(z-y) = (z-y)x by commutativity), we get

x(z−y)<0.x(z-y) < 0.

Step 3: sign of y2y^2. Since y≠0y \ne 0 (indeed y>0y > 0), (O6) gives y2>0y^2 > 0. Step 4: combine. We must add a negative term x(z−y)x(z-y) and a positive term y2y^2; the sign axioms alone (closure under ++) do not determine the sign of a sum of one positive and one negative element — that depends on magnitude, which the order axioms alone don't pin down without more numerical information. So let us make the example concrete and asymmetric: take F=QF = \mathbb{Q}, x=−3x = -3, y=1y = 1, z=5z = 5. Check hypotheses: x=−3<0x=-3<0 ✓, 0<1<50<1<5 ✓. Compute directly:

E=(−3)(5−1)+12=(−3)(4)+1=−12+1=−11<0.E = (-3)(5-1) + 1^2 = (-3)(4) + 1 = -12 + 1 = -11 < 0.

This matches Step 2's conclusion that x(z−y)<0x(z-y)<0 dominating a smaller positive term y2=1y^2=1; the sign rules correctly told us x(z−y)<0x(z-y)<0 and y2>0y^2>0 individually (that part is guaranteed by the theorem, in any ordered field, for any x,y,zx,y,z satisfying the hypotheses), while the final sign of the sum required actual numerical computation, illustrating precisely the boundary between what the order axioms guarantee structurally and what requires arithmetic. This distinction — order axioms govern signs of products and the direction of inequalities, but not automatically the sign of arbitrary sums — is itself an important lesson.

Example 2 — solving a genuine inequality using (O7) and (O8), tracking direction flips carefully. Solve for all real tt satisfying

2t−73<−4,t<0 is NOT assumed — solve in full generality.\frac{2t - 7}{3} < -4, \qquad t < 0 \text{ is NOT assumed — solve in full generality.}

Step 1. Multiply both sides by 3>03 > 0. By (O7) (first claim, multiplying by a positive element preserves direction):

2t−7<−12.2t - 7 < -12.

Step 2. Add 77 to both sides. By (O3) (translation invariance, valid for any cc, here c=7c=7):

2t<−5.2t < -5.

Step 3. Multiply both sides by 12>0\frac12 > 0. By (O7) again (positive multiplier, direction preserved):

t<−52.t < -\frac{5}{2}.

Check with a boundary-adjacent asymmetric value: let t=−3t = -3, which satisfies t<−5/2t < -5/2 since −3<−2.5-3 < -2.5. Substitute into the original expression:

2(−3)−73=−6−73=−133≈−4.333.\frac{2(-3) - 7}{3} = \frac{-6-7}{3} = \frac{-13}{3} \approx -4.333.

Indeed −4.333<−4-4.333 < -4, confirming membership in the solution set. Now check t=−2t = -2, which fails t<−5/2t < -5/2 since −2>−2.5-2 > -2.5:

2(−2)−73=−113≈−3.667,\frac{2(-2)-7}{3} = \frac{-11}{3}\approx -3.667,

and indeed −3.667≮−4-3.667 \not< -4, correctly excluded. This confirms both the direction of each step (no erroneous flip) and the final answer t<−5/2t < -5/2. A second sub-question testing (O8): for which tt in the solution set above is 1−t\frac{1}{-t} well-defined and how does it compare to 15/2\frac{1}{5/2} as t→−5/2−t \to -5/2^- from further left? Take t=−3t=-3: −t=3>0-t = 3 > 0, and 1−t=13\frac{1}{-t} = \frac13. Compare to u=5/2u = 5/2: since −t=3>5/2=u-t = 3 > 5/2 = u (as 3>2.53 > 2.5), both positive, (O8) gives

13<15/2=25.\frac{1}{3} < \frac{1}{5/2} = \frac{2}{5}.

Direct check: 1/3≈0.3331/3 \approx 0.333 and 2/5=0.42/5 = 0.4, and indeed 0.333<0.40.333 < 0.4. This confirms (O8)'s reversal rule numerically.

CounterexampleDropping trichotomy: a "positivity set" with two positive square roots of the same magnitude

The theorem's proofs lean essentially on trichotomy — without it, sign rules and even the basic dichotomy in the reciprocal proof collapse. Here is a concrete structure that satisfies closure under ++ and ×\times but fails trichotomy, and in which the sign rules of (O5) genuinely fail. Let F=Q(i)F = \mathbb{Q}(i) be the field of Gaussian rationals, {a+bi:a,b∈Q}\{a+bi : a,b\in\mathbb{Q}\}, with the usual field operations (i2=−1i^2=-1). Define

P={a+bi∈F:a>0 in Q}∪{bi:b>0 in Q},P = \{a + bi \in F : a > 0 \text{ in } \mathbb{Q}\} \cup \{bi : b > 0 \text{ in } \mathbb{Q}\},

that is, PP consists of all elements with positive real part, together with the positive-imaginary-axis elements (real part exactly 00, imaginary part positive). One checks PP is closed under addition (sum of two elements with positive real part has positive real part; the boundary case needs care but can be arranged) — but the key failure is multiplicative closure combined with trichotomy simultaneously, which is impossible in Q(i)\mathbb{Q}(i) for a much more fundamental reason: Consider a=i∈Fa = i \in F. By definition of PP above, i∈Pi \in P (it's on the positive imaginary axis, b=1>0b=1>0). If trichotomy held with this being a genuine order making FF an ordered field, (O6) would force i2>0i^2 > 0. But i2=−1i^2 = -1, and by (O6) applied to 11 (which is forced to be positive in any ordered field, as proved above), −1<0-1 < 0. So we'd need simultaneously −1>0-1 > 0 (from i2>0i^2>0) and −1<0-1<0 (from (O6) applied to 11), a contradiction. This is not a minor technicality: it is the classical theorem that the complex numbers admit no ordered field structure at all — no choice of P⊂CP \subset \mathbb{C} satisfying all three axioms exists, precisely because i2=−1i^2=-1 would force −1-1 to be both a square (hence ≥0\ge 0 by (O6)) and negative (since 1>01>0 always). This shows why trichotomy (rather than just "closure" properties) is the load-bearing hypothesis: it is what makes (O6) — and hence the sign rules — provable, and its failure (or impossibility, as in C\mathbb{C}) breaks the entire theorem.

Scope, edge cases & common traps

SituationWhat holdsWhat fails / must be checked
F=QF = \mathbb{Q} or F=RF=\mathbb{R}All of (O1)–(O8) hold exactly as proved—
F=CF = \mathbb{C}No positivity set PP exists at allOrder axioms cannot be satisfied; see counterexample
Multiplying inequality by c=0c=0ac=bc=0ac = bc = 0 trivially, no strict inequality survives(O7) explicitly requires c>0c>0 or c<0c<0, never c=0c=0
Reciprocal rule (O8)Requires both a,ba,b same sign (here positive)If a<0<ba<0<b, 1/a1/a and 1/b1/b have opposite signs; the reversal claim (O8) as stated does not apply
"Squaring preserves order"True only when both sides are non-negative−3<2-3 < 2 but (−3)2=9>4=22(-3)^2=9 > 4=2^2: squaring an inequality with a negative side can reverse or destroy it

The mathematician behind it

Origins · 1 minRichard Dedekind and how this was discovered
RD
Richard Dedekind
1831–1916 · Foundations of the real numbers, Braunschweig and Göttingen

Dedekind did not invent the order axioms in the abstract algebraic form given here — that formalization matured later, through the work of Hilbert, Hölder, and Emmy Noether's school on ordered fields and formally real fields — but Dedekind's 1872 construction of the real numbers via cuts was the decisive act that made rigorous, axiomatic treatment of order on R\mathbb{R} possible in the first place. Trained in Göttingen and later a lifelong professor in Braunschweig, Dedekind insisted that continuity and order be defined, not merely pictured, a demand that directly motivates treating "a<ba<b" as a derived relation from an axiomatically specified positivity set rather than a primitive geometric intuition.

History
- 1858/1872 — Dedekind formulates the cut construction of R\mathbb{R}, forcing a precise definition of order compatible with arithmetic.
- 1899 — Hilbert's Grundlagen der Geometrie popularizes fully axiomatic treatments of ordered structures in geometry and algebra.
- 1901 — Otto Hölder proves his theorem characterizing Archimedean ordered groups, deepening the abstract theory of order.
- 1927 — Emil Artin and Otto Schreier develop the theory of formally real fields, proving a field admits an order iff −1-1 is not a sum of squares — directly explaining why C\mathbb{C} (counterexample above) can never be ordered.
- 1930s–present — Ordered field axioms become standard foundational material in real analysis textbooks (Rudin, Spivak, Abbott), presented exactly as the starting point for building R\mathbb{R} and its inequality calculus.
Do these theorems require F=RF = \mathbb{R} specifically?

No. Every proof in this article uses only the field axioms and the three order axioms (closure of PP under ++ and ×\times, and trichotomy). It therefore holds verbatim in any ordered field: Q\mathbb{Q}, R\mathbb{R}, the field of rational functions with a suitable order, or any other structure satisfying the definition. The real numbers are distinguished from other ordered fields by an additional axiom (completeness, the least-upper-bound property), which is not used anywhere here. Everything in this article is "just order," not "specifically real numbers."

Why can't we just define a<ba < b directly and skip the positivity set PP?

You can, but then you must separately postulate all the properties you want << to have (trichotomy for <<, transitivity, compatibility with ++ and ×\times), and proving they're mutually consistent becomes circular or redundant. Defining << via a single positivity set PP with three clean axioms is more economical: trichotomy for <<, transitivity, and the sign rules all become theorems derived from three simple facts about PP, rather than a longer list of axioms about << itself that would need independent verification for consistency.

Is (−a)(−b)=ab(-a)(-b) = ab a consequence of the order axioms, or just of the field axioms?

It is a pure field-algebra identity, true in every field (ordered or not), including finite fields with no order at all. It follows from the distributive law: (−a)(−b)+(−a)b=(−a)(−b+b)=(−a)(0)=0(-a)(-b) + (-a)b = (-a)(-b+b) = (-a)(0) = 0, and separately (−a)b+ab=0(-a)b + ab = 0, so (−a)(−b)=−[(−a)b]=−[−(ab)]=ab(-a)(-b) = -[(-a)b] = -[-(ab)] = ab. What the order axioms add is not the identity itself but the extra fact that when a,ba,b are both negative (in the order sense), this common value ab=(−a)(−b)ab=(-a)(-b) is positive, which is genuinely an order-theoretic fact, not just algebra.

Why does multiplying by a negative number reverse an inequality — what's the one-line intuition?

Multiplying by c<0c<0 is the same as multiplying by −1-1 (which flips every element to its mirror image across 00, by trichotomy) and then by −c>0-c>0 (which preserves order, by the positive case). Flipping across 00 reverses which of two elements is "further right," so composing a flip with an order-preserving step yields overall reversal. The proof of (O7) makes this precise algebraically via the identity (b−a)(−c)=ac−bc∈P(b-a)(-c) = ac-bc \in P.

Does "positive times positive is positive" need its own separate axiom?

No — it literally is one of the three axioms (closure of PP under multiplication), not a derived theorem. What is derived are the mixed and doubly-negative cases: positive-times-negative-is-negative, and negative-times-negative-is-positive. Those require combining the multiplicative closure axiom with trichotomy (via the definition −a∈P-a \in P when a<0a<0) and a small amount of field algebra, as carried out explicitly in the proof of (O5).

How does this connect to absolute value, which comes next?

Absolute value is typically defined using trichotomy: ∣a∣=a|a| = a if a≥0a \ge 0 and ∣a∣=−a|a|=-a if a<0a<0, which is well-defined precisely because trichotomy guarantees every element falls into exactly one of these cases. All the standard absolute value inequalities (triangle inequality, ∣ab∣=∣a∣∣b∣|ab|=|a||b|) are then proved by casework on signs, using exactly the sign rules (O5) established here. In this sense, the current article's theorems are the direct logical prerequisites for Absolute Value and Its Properties.

Can two different positivity sets P1≠P2P_1 \ne P_2 make the same field into two genuinely different ordered fields?

Yes, in general — a field can admit multiple distinct orders (this happens for fields like Q(2)\mathbb{Q}(\sqrt2) via different embeddings into R\mathbb{R}). However, on Q\mathbb{Q} itself, and on R\mathbb{R} once completeness is added, the order is unique: it is forced to be the usual one, essentially because every positive rational (or real) must be a sum involving 11 (positive by (O6)) and closure under addition, pinning down PP completely in those specific fields.

Before this

Where this leads