Evanalysis
4.1Estimated reading time: 25 min

4.1 Total orders and ordered fields

Separate total order from partial order, then see how the familiar order on Z and Q works together with field operations.

Course contents

Why order needs its own note

Earlier chapters focused on building NN, ZZ, and QQ, then defining their operations carefully. Chapter 4 changes the viewpoint. We now ask not only what the numbers are, but also what sort of order structure they carry.

That shift matters because later concepts such as bounds, supremum, limits, and completeness all depend on the order relation. Before talking about “least upper bounds,” we need to know what kind of order we are even using.

Partial order versus total order

Definition

Total order

A set XX with relation ≤\le is totally ordered if:

  1. x≤xx\le x for every x∈Xx\in X,
  2. x≤yx\le y and y≤xy\le x imply x=yx=y,
  3. x≤yx\le y and y≤zy\le z imply x≤zx\le z,
  4. for every x,y∈Xx,y\in X, either x≤yx\le y or y≤xy\le x.

The first three conditions say that ≤\le is a partial order. The fourth condition is the extra comparability condition that makes the order total.

The crucial new feature is comparability. In a total order, there are no two elements left floating without comparison.

Worked example

A standard total order

The usual order on {1,2,3,4}\{1,2,3,4\} is total. For any two elements, one lies at or to the left of the other, so comparison is never ambiguous.

Worked example

A partial order that is not total

Let X={1,2,3,6}X=\{1,2,3,6\} and declare x≤yx\le y when xx divides yy.

This is a partial order, but not a total order. The elements 22 and 33 are incomparable: 22 does not divide 33, and 33 does not divide 22.

This example is important because it shows that “ordered” does not automatically mean “every pair is comparable.” That extra property has to be checked.

Restricted order on a subset

Once an ambient set is totally ordered, every subset inherits that order.

Theorem

Restricted order stays total

If (X,≤)(X,\le) is totally ordered and Y⊆XY\subseteq X, then YY becomes totally ordered when we use the same comparison rule on elements of YY.

Why does this matter? Because later we will study subsets such as

S={q∈Q∣q2<2}S=\{q\in Q\mid q^2\lt 2\}

inside a larger ordered set. We do not invent a new order on SS; we restrict the one already living on QQ.

The standard order on ZZ and QQ

For integers and rationals, the familiar order can be written in a way that fits the algebra already built in earlier chapters:

x≤yiff0≤y−x.x\le y \quad \text{iff} \quad 0\le y-x.

So comparison is translated into a statement about the sign of a difference. This is useful because sign, addition, and multiplication are already part of the algebraic language of ZZ and QQ.

Theorem

The standard order on ZZ and QQ is total

With the usual notion of positivity, the standard order on ZZ and on QQ is a total order.

The reason is structural: for any difference y−xy-x, exactly one of three things happens. It is positive, it is zero, or it is negative. Those three cases give x<yx\lt y, x=yx=y, or x>yx\gt y, so every pair is comparable.

Order and operations must cooperate

The order is not an isolated decoration. It has to behave properly with the field operations.

Definition

Order-compatible algebra facts

For rational numbers x,y,zx,y,z:

  • if x≤yx\le y, then x+z≤y+zx+z\le y+z;
  • if x≥0x\ge 0 and y≥0y\ge 0, then xy≥0xy\ge 0.

The first rule says translation preserves order. The second says multiplying nonnegative quantities cannot suddenly create a negative result.

Worked example

Why translation invariance matters

If 1/3≤1/21/3\le 1/2, then adding 22 to both sides gives

2+13≤2+12.2+\frac13 \le 2+\frac12.

This is not a new theorem each time. It is the same structural rule applied to different numbers.

Without these compatibility rules, the order would not interact correctly with the algebra. Later arguments about intervals, bounds, and absolute values would fall apart.

Fields and ordered fields

One can summarize almost everything we know about QQ by packaging the operations together.

Definition

Field

A field is a set FF with distinct distinguished elements 0,1∈F0,1\in F, addition and multiplication maps F×F→FF\times F\to F, and the following laws:

  • addition and multiplication are associative and commutative,
  • 0≠10\ne1, and 0+x=x0+x=x, 1x=x1x=x for all x∈Fx\in F,
  • multiplication distributes over addition,
  • every xx has an additive inverse −x-x with x+(−x)=0x+(-x)=0,
  • every x≠0x\ne0 has a multiplicative inverse x−1x^{-1} with xx−1=1xx^{-1}=1.

Closure is part of the operation-map statement: adding or multiplying two elements of FF must return an element of FF.

This definition remembers the algebra, but it says nothing about order.

Definition

Ordered field

An ordered field is a field equipped with a total order ≤\le such that:

  • x≤yx\le y implies x+z≤y+zx+z\le y+z,
  • x≥0x\ge 0 and y≥0y\ge 0 imply xy≥0xy\ge 0.

So an ordered field is not just “a field plus a comparison symbol.” It is a field whose algebra and order fit together coherently.

The rational numbers QQ are an ordered field. The real numbers RR will also be an ordered field, but chapter 4 will show that RR has one more crucial property beyond this: completeness.

Common mistake

A total order is more than a left-to-right picture

Students often think a total order is just “something that can be drawn on a line.” The real point is the comparability axiom. A partial order can still have a clear diagram or hierarchy and yet fail to compare some pairs.

Common mistake

A field is not automatically an ordered field

The field axioms only describe algebraic operations. To become an ordered field, the set also needs a total order, and that order must respect addition and multiplication in the precise way stated above.

Comparing orders by the hypotheses they support

There are two questions that are often compressed into the word “order.” The first asks whether two elements can be compared. The second asks whether that comparison is compatible with arithmetic. The first is answered by a relation; the second requires extra structure.

Definition

Strict order and non-strict order

In a partially ordered set, write x<yx\lt y when x≤yx\le y and x≠yx\ne y. In a total order, exactly one of x<yx\lt y, x=yx=y, or y<xy\lt x holds for every pair. In an ordered field, the strict relation is also compatible with translation and positive scaling: x<yx\lt y implies x+z<y+zx+z\lt y+z, and 0<z0\lt z implies xz<yzxz\lt yz.

The distinction explains why the same-looking argument can be valid in QQ but invalid in a set ordered by divisibility. In QQ, the expression (y−x)/2(y-x)/2 is available and positivity can be transported through it. In a divisibility order, “halfway between” two elements is not an order-theoretic notion at all.

Worked example

A translation argument in an ordered field

Assume a<ba\lt b and c<dc\lt d in an ordered field. Adding cc to the first inequality gives a+c<b+ca+c\lt b+c; adding bb to the second gives b+c<b+db+c\lt b+d. Transitivity yields a+c<b+da+c\lt b+d. Notice the proof uses only translation invariance and transitivity; it does not rely on a picture of the number line.

Worked example

A positive-product argument

If 0<a<b0\lt a\lt b and 0<c<d0\lt c\lt d, then ac<bcac\lt bc because multiplying by the positive cc preserves strict order, and bc<bdbc\lt bd because multiplying by the positive bb preserves the second inequality. Thus ac<bdac\lt bd. The sign hypotheses are not cosmetic: multiplying by a negative number reverses an inequality.

The order laws in proof form

The definitions become useful only when we can turn them into reliable proof steps. For x,y∈Qx,y\in Q, exactly one of y−x>0y-x\gt0, y−x=0y-x=0, or y−x<0y-x\lt0 holds. Using x≤yx\le y precisely when 0≤y−x0\le y-x, these three cases give x<yx\lt y, x=yx=y, or y<xy\lt x. This proves trichotomy and hence comparability for the standard order. It also shows why the order is antisymmetric: if both x≤yx\le y and y≤xy\le x, then both y−xy-x and x−yx-y are nonnegative, so their sum is zero and therefore x=yx=y.

Here is the translation proof in its general form. If x≤yx\le y, then 0≤y−x0\le y-x. For any zz,

(y+z)−(x+z)=y−x,(y+z)-(x+z)=y-x,

so the same nonnegative difference gives x+z≤y+zx+z\le y+z. Conversely, applying this rule with −z-z shows that x+z≤y+zx+z\le y+z implies x≤yx\le y; translation is therefore an order isomorphism. In particular, subtraction is not a separate order rule: it is translation by an additive inverse.

Positive scaling requires a sign hypothesis. If x≤yx\le y and 0≤z0\le z, then 0≤y−x0\le y-x and multiplication of nonnegative rationals gives 0≤z(y−x)=zy−zx0\le z(y-x)=zy-zx; hence zx≤zyzx\le zy. If z>0z>0 and x<yx<y, then y−x>0y-x>0, so z(y−x)>0z(y-x)>0 and the inequality is strict. If z<0z<0, write z=−wz=-w with w>0w>0; positive scaling by ww followed by negation reverses the inequality. This is the algebraic reason that cancellation by a positive number preserves order, whereas cancellation by a negative number reverses it.

Theorem

Trichotomy and order compatibility in Q

For every x,y∈Qx,y\in Q, exactly one of x<yx\lt y, x=yx=y, and y<xy\lt x holds. Moreover, translation by any rational preserves both ≤\le and <<, while multiplication by a positive rational preserves them and multiplication by a negative rational reverses them.

These statements also clarify the quantifiers used later. The assertion x≤yx\le y is a statement about one pair. The assertion that uu is an upper bound of YY is stronger:

(∀y∈Y)  y≤u.(\forall y\in Y)\;y\le u.

To refute it, one must produce a single witness y∈Yy\in Y with u<yu\lt y in a total order. A picture with several points below uu cannot establish the universal claim, and a list of examples cannot refute it without such a witness. The empty set is a useful boundary case: every uu satisfies the upper-bound condition for ∅\varnothing vacuously, because there is no element to test. An unbounded set behaves differently: QQ itself has no upper bound, since for any proposed u∈Qu\in Q, the rational u+1u+1 lies above it.

Worked example

Three subsets, three different order questions

In X={−2,0,3}X=\{-2,0,3\} with its usual total order, Y={−2,3}Y=\{-2,3\} has minimum −2-2 and maximum 33; its upper bounds in XX are just 33, while its upper bounds in QQ are every rational u≥3u\ge3. The interval Z=(0,1)Z=(0,1) has no maximum, although 11 is an upper bound in QQ; for every u<1u\lt1, a point of ZZ lies above uu. Finally, QQ is unbounded above, because u+1u+1 defeats every candidate uu. The ambient ordered set and the quantifiers change the answer.

For example, an interval such as [0,1]⊆R[0,1]\subseteq R inherits a total order, but it is not a field: 22 is not in the interval, and a positive element such as 1/21/2 has no additive inverse in the interval. We may discuss bounds and suprema of subsets of the interval, but we must name RR as the ambient ordered field when invoking completeness or subtraction.

Common mistake

Closure is separate from inherited order

Every subset inherits comparisons, but it need not inherit algebraic closure. Do not call an interval a field merely because it sits inside RR; state the ambient field whenever an argument uses its operations.

A compact proof toolkit for ordered fields

Several consequences of the ordered-field axioms recur so often that they are worth proving once. The multiplicative axiom initially gives only a≥0a\ge0 and b≥0b\ge0 implying ab≥0ab\ge0; strict positivity must be derived. First prove 0<10\lt1. The field axiom gives 1≠01\ne0. If 1<01\lt0, translation by −1-1 gives 0<−10\lt-1, so −1≥0-1\ge0. The nonnegative-product axiom then gives

0≤(−1)2=1,0\le(-1)^2=1,

contradicting 1<01\lt0. Trichotomy therefore gives 0<10\lt1. If a<ba\lt b, adding −a-a gives 0<b−a0\lt b-a, so every strict inequality can be rewritten as positivity of a difference.

Now suppose 0<a0\lt a and 0<b0\lt b. The axiom gives ab≥0ab\ge0. A field has no zero divisors: if ab=0ab=0, then multiplying by a−1a^{-1} or b−1b^{-1} would force one of a,ba,b to be zero. Hence ab≠0ab\ne0, and trichotomy upgrades ab≥0ab\ge0 to 0<ab0\lt ab. In particular, if a≠0a\ne0, then a2≥0a^2\ge0 and a2≠0a^2\ne0, so a2>0a^2\gt0. This gives the useful implication

a2=0⟹a=0.a^2=0\quad\Longrightarrow\quad a=0.

The inverse of a positive element is positive as well. If c>0c\gt0, then c−1≠0c^{-1}\ne0 because cc−1=1cc^{-1}=1. It cannot be negative: if c−1<0c^{-1}\lt0, then −c−1>0-c^{-1}\gt0, and the strict-product result would give c(−c−1)=−1>0c(-c^{-1})=-1\gt0, contradicting −1<0-1\lt0. Thus c−1>0c^{-1}\gt0. Now cancellation is justified: if c>0c\gt0 and ac≤bcac\le bc, then 0≤(b−a)c0\le(b-a)c; multiplying by the positive inverse c−1c^{-1} preserves nonnegativity and gives 0≤b−a0\le b-a, hence a≤ba\le b. This is why inequalities involving fractions must record that their denominator is positive. For example, a/b<c/da/b<c/d may be cross-multiplied without reversing direction only after checking b>0b>0 and d>0d>0.

Worked example

Checking an inequality by its difference

To prove −2/3<5/6-2/3\lt5/6, do not rely on a diagram. Compute

56−(−23)=56+46=96=32>0.\frac56-\left(-\frac23\right)=\frac56+\frac46=\frac96=\frac32\gt0.

The positivity of this difference is exactly the definition of the strict order. If both sides are translated by 77, the difference remains 3/23/2, so 19/3<47/619/3\lt47/6 follows from the same proof.

These elementary implications are the bridge from the abstract axioms to later bound arguments. They justify taking midpoints such as (a+b)/2(a+b)/2 in an ordered field, but they do not justify taking a midpoint in an arbitrary partial order. The distinction will matter when a supremum argument is transferred from a total order to a partial order.

Quantifiers determine the order claim

Order notation often hides the domain over which a statement is quantified. The claim x<yx\lt y is local, but an interval statement is universal. For example,

(∀x∈(0,1))  x<1(\forall x\in(0,1))\;x<1

proves that 11 is an upper bound of (0,1)(0,1), while

(∀u<1)(∃x∈(0,1))  u<x(\forall u<1)(\exists x\in(0,1))\;u<x

proves that no smaller real number is an upper bound. The order alone does not turn one of these formulas into the other; the second needs an explicit witness, such as x=(u+1)/2x=(u+1)/2 when u≥0u\ge0, or x=1/2x=1/2 when u<0u<0.

The same discipline handles lower bounds. To prove that 00 is the infimum of (0,1)(0,1), show 0≤x0\le x for every xx in the interval, then take any v>0v>0 and choose a point of the interval below vv. The choice x=min⁡{v/2,1/2}x=\min\{v/2,1/2\} works because it is positive, below 11, and strictly below vv. These witness formulas are the elementary form of the approximation criteria developed in the next note.

Empty and unbounded domains expose the limits of the definitions. For Y=∅Y=\varnothing, every uu is both an upper and a lower bound, since a universal statement over no elements is true. Thus a claim about a “best” bound for the empty set needs a separate convention and is excluded from the usual completeness axiom. For Y=QY=Q, every proposed upper bound uu is defeated by u+1u+1, and every proposed lower bound by u−1u-1; the set is unbounded in both directions. The quantifiers are what make these conclusions rigorous.

Worked example

One interval, four quantified statements

For Y=[−1,2]Y=[-1,2] in RR, 22 is an upper bound because every y∈Yy\in Y satisfies y≤2y\le2, and it is a maximum because 2∈Y2\in Y. The element 33 is also an upper bound but is not a maximum because 3∉Y3\notin Y. The number −1-1 is a minimum and an infimum. If the interval is changed to (−1,2)(-1,2), the same bound inequalities remain true, but neither endpoint is a member; therefore there is no maximum or minimum. Changing one strict symbol changes membership without changing the surrounding order structure.

These examples illustrate a general proof habit. Begin by naming the ambient ordered set, because the same subset can have different available bounds after the universe changes. Next write the statement with its quantifiers. A maximum claim contains an existential membership clause and a universal comparison clause; an upper-bound claim contains only the universal comparison clause. Finally, when proving that a bound is sharp, state how a proposed improvement is defeated. For an upper bound this means finding a set element above the proposal; for a lower bound it means finding a set element below it.

The algebraic axioms support these witnesses in an ordered field. Midpoints are available because division by 22 is defined and 2>02>0. Positive scaling allows an inequality to be transported through multiplication, and additive inverses allow every translation argument to be reversed. None of these facts creates a supremum. They only let us verify a candidate once one has been identified. The existence of candidates for all nonempty bounded sets is the separate completeness property introduced in the next note.

This separation is the reason the chapter begins with order before discussing real-number construction. A construction must eventually supply a total order, an ordered-field algebra, and a completeness principle. If one verifies only the algebra, a field can still have incomparable elements or fail to contain a least upper bound. If one verifies only the order, a subset can inherit comparisons while losing closure under addition, multiplication, or inverses. The three layers should remain visible in every later proof.

One practical consequence is that an inequality proof should always identify its permitted operations. Translation is valid for every field element, while division requires a nonzero denominator and preserves direction only for a positive denominator. Multiplication preserves a weak inequality when the factor is nonnegative, and preserves a strict inequality when the factor is positive. If the factor is negative, the direction reverses. These are not stylistic qualifications: omitting one can change a true statement into a false one.

The order axioms also explain why intervals are natural test cases. An interval inherits the ambient comparison, so one can ask for its bounds. It generally does not inherit the field operations, so one must perform arithmetic in the ambient field. This separation lets us use a small interval to test the difference between membership and comparison before confronting infinite sets and missing endpoints.

Quick checks

Checkpoint

Why is divisibility on positive integers not a total order?

Look for two incomparable elements.

Solution · Answer

Because some pairs cannot be compared. For example, 22 does not divide 33, and 33 does not divide 22, so the comparability condition fails.

Checkpoint

If x≤yx\le y, why does x−z≤y−zx-z\le y-z also hold?

Rewrite subtraction as addition.

Solution · Answer

Because x−z=x+(−z)x-z = x+(-z) and y−z=y+(−z)y-z = y+(-z). The order-compatibility rule for addition says adding the same number to both sides preserves the inequality.

Exercises

Checkpoint

Explain why every subset of a totally ordered set inherits a total order.

Take two elements of the subset and compare them in the ambient set.

Solution · Guided solution

Let y1,y2∈Y⊆Xy_1,y_2\in Y\subseteq X. Since XX is totally ordered, either y1≤y2y_1\le y_2 or y2≤y1y_2\le y_1 in XX. The restricted order on YY uses the same comparison, so the same alternative holds in YY. The other order axioms were already true in XX, so they remain true on the subset.

Checkpoint

Why is Q>0Q_{\gt 0} not a field?

Check closure and inverses against the field axioms.

Solution · Guided solution

The positive rationals are closed under multiplication and have multiplicative inverses, but they do not have additive inverses inside the same set. If q>0q\gt 0, then −q-q is not in Q>0Q_{\gt 0}. So the field axioms fail.

Read this after 3.4 Rationals and well-defined operations and 3.5 Gaps in Q and sqrt(2). Then continue to 4.2 Upper bounds, supremum, and infimum.

Practice

Work out your answer, then check it. You can revise and try again.

Loading…

Key terms in this unit