6.2  Building the Map from the Rationals Below

Section 6.1 gave every element x of an ordered field F a record, the set QF(x) of rationals below it, and showed that in an Archimedean field the record determines the element (Lemma 6.1.3). This section turns the record into a map. Given an Archimedean field F and a complete field L, we send x∈F to the element of L with the same record. The result is the main theorem of the chapter in its one-sided form: there is exactly one order-preserving field map F→L. Section 6.3 applies it twice, in both directions, to compare two complete ordered fields.

Two fields are in play throughout, so every proof here keeps ιF and ιL visible. Only the two lemmas, each about one field, write q for ιF(q).

6.2.1. What the map must be

Suppose φ:F→L is an order-preserving field map (Definition 6.1.1). By Proposition 6.1.1 it fixes the rationals, φ(ιF(q))=ιL(q), and it preserves and reflects order. So for q∈Q and x∈F, ιF(q)<x⟺φ(ιF(q))<φ(x)⟺ιL(q)<φ(x). In the notation of Section 6.1: QL(φ(x))=QF(x). The image of x has exactly the rationals below it that x has.

That already shows there can be at most one such map when L is Archimedean: two candidates for φ(x) would have the same rationals below them, and rationals separate. For existence, we need an element of L with a prescribed set of rationals below it. When L is complete there is a natural candidate: the least element of L that lies above all the copies ιL(q), q∈QF(x), that is, their supremum.

The check that this candidate respects sums and products goes through the record, not through suprema. For that we need to know how the rationals below a+b and below ab are made from those below a and b.

6.2.2. Sums and products, measured by rationals

A rational q lies below a+b exactly when it splits into a rational below a and a rational below b.

Lemma 6.2.1 (Rationals below a Sum)

Let F be an Archimedean ordered field, a,b∈F and q∈Q. Then ιF(q)<a+b if and only if there is r∈Q with ιF(r)<aandιF(q−r)<b. In other words, q∈QF(a+b)⟺∃r∈Q (r∈QF(a)∧q−r∈QF(b)).

Idea. One direction is addition of inequalities. For the other, split q as r+(q−r) with r a rational just below a: we need r in the gap between q−b and a, which is non-empty exactly because q<a+b, and density supplies it.

Proof. We write q for ιF(q); since ιF preserves sums, it preserves differences, so ιF(q−r)=q−r in F.

(⇐) If r<a and q−r<b, then q=r+(q−r)<a+b by Proposition 2.3.5 (b).

(⇒) Let q<a+b. Then q−b<a (Proposition 2.3.5 (a)). By Theorem 4.2.5 there is r∈Q with q−b<r<a. Then r<a, and q−r<b by Proposition 2.3.5 (a) again.

For products the same idea works for positive elements and positive rationals, with q/r in place of q−r.

Lemma 6.2.2 (Rationals below a Product)

Let F be an Archimedean ordered field, a,b∈F with a>0 and b>0, and q∈Q with q>0. Then ιF(q)<ab if and only if there is r∈Q with r>0,ιF(r)<aandιF(q/r)<b. For q≤0, ιF(q)<ab holds always.

In words: below a product of positives, a positive rational splits as a product of a rational below each factor. Exercise B1 asks for the proof; the fold below has it.

Second readingThe proof of the product lemma (try Exercise B1 first)

Idea. As for sums, with division in place of subtraction: split q as r⋅(q/r) with r a rational in the gap between q/b and a, a gap of positive elements.

Proof. We write q for ιF(q); since ιF preserves products and ιF(r)≠0 for r≠0, it preserves quotients, so ιF(q/r)=q/r in F.

If q≤0, then q≤0<ab by (O4). Let q>0.

(⇐) Let r>0 with r<a and q/r<b. Since q/r>0 (Proposition 2.5.2 (a), (O4)), Proposition 2.3.5 (c) gives q=r⋅(q/r)<a⋅(q/r); since a>0, it gives a⋅(q/r)<ab. So q<ab.

(⇒) Let q<ab. Multiplying by b−1>0 (Proposition 2.5.2 (a), Proposition 2.3.5 (c)) gives q/b<a, and q/b>0 by (O4). By Theorem 4.2.5 there is r∈Q with q/b<r<a. Then r>q/b>0, so r>0 in Q, since ιF reflects order. Multiplying q/b<r by b>0 gives q<rb, and multiplying by r−1>0 gives q/r<b.

Quick check. In F=Q, with a:=3/4, b:=1/2 and q:=1, find a rational r as in Lemma 6.2.1. Which rationals r work?

Solution. We need r<3/4 and 1−r<1/2, that is, 1/2<r<3/4. For example r=2/3: then 1−2/3=1/3<1/2. Exactly the rationals of the open gap between q−b=1/2 and a=3/4 work, as the proof predicts.

6.2.3. The map into a complete field

Here is the theorem. Its proof is the first key proof of the chapter.

Theorem 6.2.3 (Embedding an Archimedean Field in a Complete One)

Let F be an Archimedean ordered field and L a complete ordered field. Then there is exactly one order-preserving field map φ:F→L. It is given by φ(x):=supL{ιL(q):q∈QF(x)}(x∈F), the supremum taken in L, and it satisfies (⋆)QL(φ(x))=QF(x)for every x∈F.

Try it first (ten minutes): if φ exists and fixes Q, which element of L must φ(x) be?

Idea. The natural first attempt defines φ on the copy of Q, by φ(ιF(q)):=ιL(q), and then tries to “extend”. It stalls at once: an element x∈F that is not a rational has no formula, and nothing in F names it. What we do know, from the previous subsection, is its record: φ(x) must have exactly the rationals QF(x) below it. So we construct a point as a supremum: φ(x) is the least upper bound in L of the copies of those rationals, and (⋆) says the construction has the right record. After that, sums, products and uniqueness are checked the same way: compute the record of both sides (for sums and products by the two lemmas) and let rationals separate in L. Monotonicity needs only one rational between x and y.

Step roadmap: ① φ(x) exists; ② (⋆); ③ φ fixes the rationals; ④ φ is strictly increasing; ⑤ sums; ⑥ products and the unit; ⑦ uniqueness. Before reading on: where is completeness of L used, and where the Archimedean property of F?

Proof. By Theorem 6.1.2, L is Archimedean. So Theorem 4.2.5, Lemma 6.1.3, Lemma 6.2.1 and Lemma 6.2.2 apply in L as well as in F. For x∈F write Ax:={ιL(q):q∈QF(x)}⊆L.

Step 1: φ(x) is defined. Let x∈F. Since F is Archimedean, there is n∈N with −x<n⋅1F, so ιF(−n)=−(n⋅1F)<x (Proposition 2.3.5 (a)), and −n∈QF(x): the set Ax is non-empty. Applied to x, the Archimedean property gives m∈N with x<m⋅1F=ιF(m). For q∈QF(x), ιF(q)<x<ιF(m), so q<m because ιF reflects order, and ιL(q)<ιL(m). So ιL(m) is an upper bound of Ax. Since L is complete, Ax has a supremum in L, and it is unique; φ(x) is that supremum. So φ:F→L is a function.

Step 2: (⋆). Let x∈F and q∈Q.

(⊆) Let q∈QF(x). By density in F there is r∈Q with ιF(q)<ιF(r)<x. Then r∈QF(x), so ιL(r)∈Ax and ιL(r)≤φ(x), since φ(x) is an upper bound. Also q<r (reflection in F), so ιL(q)<ιL(r)≤φ(x), and q∈QL(φ(x)).

(⊇) Let ιL(q)<φ(x). Since φ(x) is the least upper bound of Ax, ιL(q) is not an upper bound, so there is r∈QF(x) with ιL(q)<ιL(r) (totality). Then q<r (reflection in L), so ιF(q)<ιF(r)<x, and q∈QF(x).

Step 3: φ fixes the rationals. Let r∈Q. By (⋆) and the first example after Definition 6.1.2, used in F and in L, QL(φ(ιF(r)))=QF(ιF(r))={q∈Q:q<r}=QL(ιL(r)). By Lemma 6.1.3 in L, φ(ιF(r))=ιL(r). In particular φ(0F)=0L and φ(1F)=1L.

Step 4: φ is strictly increasing. Let x<y in F. By density in F there is r∈Q with x<ιF(r)<y. Then r∉QF(x) (by (O1)), so r∉QL(φ(x)) by (⋆), and φ(x)≤ιL(r) by totality. And r∈QF(y)=QL(φ(y)), so ιL(r)<φ(y). Hence φ(x)≤ιL(r)<φ(y).

Step 5: sums. Let x,y∈F and q∈Q. Then q∈QL(φ(x+y))⟺q∈QF(x+y)⟺∃r∈Q (r∈QF(x)∧q−r∈QF(y))⟺∃r∈Q (r∈QL(φ(x))∧q−r∈QL(φ(y)))⟺q∈QL(φ(x)+φ(y)), by (⋆), Lemma 6.2.1 in F, (⋆) twice, and Lemma 6.2.1 in L. By Lemma 6.1.3 in L, φ(x+y)=φ(x)+φ(y).

Step 6: products and the unit. First let x>0 and y>0. By Steps 3 and 4, φ(x)>φ(0F)=0L and φ(y)>0L; and xy>0, φ(x)φ(y)>0 by (O4). Let q∈Q. If q≤0, then q∈QF(xy)=QL(φ(xy)), and q∈QL(φ(x)φ(y)), by the last clause of Lemma 6.2.2 in F and in L. If q>0, then q∈QL(φ(xy))⟺q∈QF(xy)⟺∃r∈Q>0 (r∈QF(x)∧q/r∈QF(y))⟺∃r∈Q>0 (r∈QL(φ(x))∧q/r∈QL(φ(y)))⟺q∈QL(φ(x)φ(y)), by (⋆), Lemma 6.2.2 in F, (⋆) twice, and Lemma 6.2.2 in L (with φ(x),φ(y)>0). By Lemma 6.1.3 in L, φ(xy)=φ(x)φ(y) for x,y>0.

Now the signs. By Steps 5 and 3, φ(x)+φ(−x)=φ(0F)=0L, so φ(−x)=−φ(x) (Proposition 2.3.2 (a)). If x=0 or y=0, both sides of φ(xy)=φ(x)φ(y) are 0L (Step 3 and Proposition 2.3.2 (b)). If x<0<y, then −x>0 (Proposition 2.3.5 (a)) and xy=−((−x)y) (Proposition 2.3.2 (c)), so φ(xy)=−φ((−x)y)=−(φ(−x)φ(y))=−(−φ(x))φ(y)=φ(x)φ(y). The case x>0>y follows by (F6), and if x,y<0, then xy=(−x)(−y) and φ(xy)=(−φ(x))(−φ(y))=φ(x)φ(y). Finally φ(1F)=1L by Step 3. With Steps 4 and 5, φ is an order-preserving field map.

Step 7: uniqueness. Let χ:F→L be any order-preserving field map, and let x∈F, q∈Q. By Proposition 6.1.1, χ(ιF(q))=ιL(q). If ιF(q)<x, then ιL(q)=χ(ιF(q))<χ(x). If ιL(q)<χ(x), that is, χ(ιF(q))<χ(x), then ιF(q)<x, because χ reflects order (the remark after Definition 6.1.1). So QL(χ(x))=QF(x)=QL(φ(x)), the second equality by (⋆). By Lemma 6.1.3 in L, χ(x)=φ(x). Since x was arbitrary, χ=φ. This proves the theorem.

The answer to the prompt. Completeness of L is used in Step 1, where the supremum must exist, and in Step 2 (⊇), where its leastness turns “below φ(x)” into “below some ιL(r)”. The Archimedean property of F is used in Step 1 for the bounds −n and m, in Steps 2 and 4 through density in F, and in Steps 5 and 6 through the two lemmas applied in F. Completeness of L is used a third time, indirectly: it makes L Archimedean (Theorem 6.1.2), which Steps 3, 5, 6 and 7 need for rationals to separate in L and for the lemmas in L.

Remark. Step 7 is the whole of “uniqueness”. Its idea, comparing the rationals below two candidate values, needs only that L is Archimedean, not that it is complete: Exercise C1 runs it for any two order-preserving field maps F→L, with no appeal to (⋆). Section 6.3 leans on the uniqueness clause throughout: to show that the maps between two complete fields are inverse to each other, that the isomorphism they give is the only one, and that a complete ordered field has no order-preserving field automorphism other than the identity.

Remark. Neither hypothesis can be dropped from the existence half. Without completeness of L: take F=R, which is Archimedean (Theorem 5.4.2), and L=Q. An order-preserving field map φ:R→Q would fix 2 (Proposition 6.1.1), so for 2∈R (the element of Example (The Class of the Newton Sequence), as in Section 6.1), φ(2)2=φ((2)2)=φ(2)=2 in Q, against Theorem 2.6.1. Without the Archimedean property of F: take F=Q(t) of Example (A Non-Archimedean Ordered Field) and any complete L. An order-preserving field map φ would fix each n⋅1 and be strictly increasing, so n⋅1L=φ(n⋅1)<φ(t) for every n∈N, against the Archimedean property of L (Theorem 6.1.2).

Warning. The supremum in the definition of φ is taken in L, of the copies ιL(q). It cannot be taken in Q, or in F. In Q: the set QR(2) of Example (The Rationals below the Square Root of Two) (with 2 as in Section 6.1) has no supremum in Q. It contains the set A={p∈Q:p2<2} of Theorem 2.6.2, and its other elements are negative, so below 0∈A; hence the two sets have the same upper bounds in Q, and A has no least one. In F: the supremum of ιF(QF(x)) always exists and is x itself, since x is an upper bound and, by density in F (Theorem 4.2.5), every u<x lies below some ιF(q)<x. So it is an element of the wrong field.

Quick check. Take L:=F, a complete ordered field. Which map F→F does Theorem 6.2.3 produce, and why, without computing a single supremum?

Solution. The identity. The theorem says there is exactly one order-preserving field map F→F (F is Archimedean by Theorem 6.1.2), and idF is one (the first example after Definition 6.1.1). So φ=idF. This is the argument Section 6.3 builds on.

6.2.4. Exercises

A. Check your understanding

Exercise (A1)

These questions are about the proof of Theorem 6.2.3.

  • (a) Which steps use that L is complete, and which that F is Archimedean? Which steps use that L is Archimedean, and where does that come from?
  • (b) Run the definition of φ with F=Q and L=R, the constructed reals. What is φ?
  • (c) State the idea of the proof in two sentences.
  • (d) Why may Step 1 not define φ(x) as a supremum taken in F?

Solution.

  • (a) L complete: Step 1 (the supremum exists) and Step 2 (⊇) (its leastness). F Archimedean: Step 1 (the bounds −n, m), Steps 2 and 4 (density in F), and Steps 5 and 6 (Lemma 6.2.1 and Lemma 6.2.2 applied in F). L Archimedean: Steps 3, 5, 6 and 7, through Lemma 6.1.3 in L and the two lemmas in L. It comes from the completeness of L, by Theorem 6.1.2.

  • (b) φ=ιR=κ. By Step 3, φ(ιQ(r))=ιR(r) for every r∈Q, and ιQ is the identity (Chapter 4, after Theorem 4.2.2). So φ(r)=ιR(r), which is κ(r) by Theorem 5.4.1 (d). Concretely, φ(r) is the supremum in R of {κ(q):q<r}, and that supremum is κ(r).

  • (c) An order-preserving map that fixes the rationals must send x to an element with the same rationals below it, and in a complete L the supremum of their copies is such an element. Every field identity for φ is then checked by comparing the rationals below the two sides, which the sum and product lemmas compute, and rationals separate in L.

  • (d) In F the supremum of ιF(QF(x)) always exists, since F is Archimedean, and it is x itself (Exercise B2 of Section 6.1): an element of F, not of L. So it does not give a map into L. The supremum must be of the copies ιL(q), taken in L.

B. Practice

Exercise (B1)

Prove Lemma 6.2.2: let F be an Archimedean ordered field, a,b∈F with a,b>0, and q∈Q with q>0; show that ιF(q)<ab if and only if some rational r>0 has ιF(r)<a and ιF(q/r)<b. Show also that ιF(q)<ab for every rational q≤0.

Hint: imitate the proof of Lemma 6.2.1, with division in place of subtraction.

Solution. Write q for ιF(q). Since ιF preserves products, ιF(r)ιF(r−1)=ιF(1)=1 for r≠0, so ιF(r−1)=ιF(r)−1 (Lemma 4.1.1 (a)) and ιF(q/r)=q/r in F.

If q≤0, then q≤0<ab by (O4).

Let q>0. (⇐) Let r>0, r<a, q/r<b. Then q/r>0 (Proposition 2.5.2 (a), (O4)), so Proposition 2.3.5 (c) gives q=r(q/r)<a(q/r), and, as a>0, a(q/r)<ab. Hence q<ab.

(⇒) Let q<ab. Since b−1>0, Proposition 2.3.5 (c) gives q/b<a, and q/b>0. By Theorem 4.2.5 there is r∈Q with q/b<ιF(r)<a. Then ιF(r)>0=ιF(0), so r>0 since ιF reflects order. From q/b<r and b>0, q<rb; multiplying by r−1>0, q/r<b. This r is as required.

Exercise (B2: Cube roots)

Let F be a complete ordered field and a∈F with a>0. Let S:={y∈F:y≥0 and y3<a}. The following steps prove that some s∈F has s3=a.

  • (1) S is non-empty (0∈S) and bounded above by 1+a: if y>1+a, then y>1, so y3>y>a. Let s:=supS. Then s>0, since m:=min{1,a/2} has 0<m≤1, so m3≤m≤a/2<a, and m∈S.
  • (2) For y,z∈F, z3−y3=(z−y)(z2+zy+y2) and z2+zy+y2=(z+y/2)2+(3/4)y2, which is >0 unless y=z=0 (if y≠0, then (3/4)y2>0 by Proposition 2.3.5 (e), (c) and the square is ≥0, so the sum is positive by (b); if y=0 and z≠0, the expression is z2>0 by (e)). Hence y<z⟺y3<z3, for all y,z∈F.
  • (3) The case s3<a is impossible. Let h:=min{1,(a−s3)/(2(3s2+3s+1))}>0. Since 0<h≤1 and s≥0, (s+h)3=s3+3s2h+3sh2+h3≤s3+h(3s2+3s+1)≤s3+(a−s3)/2<a, so s+h∈S, against s being an upper bound.
  • (4) The case s3>a is impossible. Name the move, then complete it.
  • (5) Hence s3=a.

Then explain how to get a cube root of a=0 and of a<0, and why every a∈F has exactly one cube root.

Solution. The move is construct a point as a supremum: s is built as supS, and each of the two wrong cases of trichotomy is refuted by exhibiting a point that contradicts one clause of the supremum. In step (3) it was an element of S above s, against (S1). In step (4) it is an upper bound below s, against (S2).

  1. Suppose s3>a. Let h:=(s3−a)/(3s2); it is positive, since s>0 by (1). Also h<s: this says s3−a<3s3, which holds because a>0 and 2s3>0. So s−h>0, and (s−h)3−a=(s3−a)−3s2h+3sh2−h3=h2(3s−h)>0, since 3s2h=s3−a, h2>0 and 3s−h>2s>0. Let y>s−h. By (2), y3>(s−h)3>a, so y∉S. Hence every element of S is ≤s−h, and s−h is an upper bound of S smaller than s, against s being the least upper bound.

  2. Steps (3) and (4) leave only s3=a, by trichotomy (O1).

For a=0, 03=0. For a<0, −a>0 has a cube root c by the above, and (−c)3=−c3=a (Proposition 2.3.2 (c)). Uniqueness: if b3=c3, then neither b<c nor c<b can hold, since by (2) either would make the cubes differ; so b=c. So every element of a complete ordered field has exactly one cube root.

C. Going deeper

Exercise (C1: Exit check: at most one map into an Archimedean field)

Let F and L be Archimedean ordered fields, and let χ1,χ2:F→L be order-preserving field maps. Prove that χ1=χ2.

Solution (Hint 1). Compare the rationals below χ1(x) and below χ2(x).

Solution (Hint 2). By Proposition 6.1.1, χi(ιF(q))=ιL(q) for i=1,2 and every q∈Q.

Solution. Let x∈F and i∈{1,2}. By Proposition 6.1.1 (which needs no order hypothesis), χi(ιF(q))=ιL(q) for every q∈Q. Since χi is strictly increasing and reflects order (the remark after Definition 6.1.1), ιF(q)<x⟺χi(ιF(q))<χi(x)⟺ιL(q)<χi(x). So QL(χ1(x))=QF(x)=QL(χ2(x)). Since L is Archimedean, Lemma 6.1.3 gives χ1(x)=χ2(x). As x was arbitrary, χ1=χ2. (This is Step 7 of Theorem 6.2.3 with χ2 in place of φ; completeness of L is not needed, and neither is the Archimedean property of F.)

Why L must be Archimedean (remark, not proved here). The last step separates χ1(x) from χ2(x) by a rational strictly between them, which is density of ιL(Q) in L, and density needs Archimedes (Theorem 4.2.5). In the non-Archimedean field Q(t) of Example (A Non-Archimedean Ordered Field), the elements 1/t and 2/t are different and have the same rationals below them, namely all q≤0: a positive rational q has q>1/n in Q for some n∈Z>0 (Theorem 2.5.3), so ιQ(t)(q)>ιQ(t)(1/n)=1/(n⋅1)>2/t, because t>2n⋅1 (Example (A Non-Archimedean Ordered Field)). So the separating step fails in such a field. For the stronger statement proved above, with F arbitrary, the hypothesis on L cannot be dropped: on Q(t), the substitution t↦t+1 keeps degrees and leading coefficients, so it is an order-preserving field map Q(t)→Q(t) different from the identity (school polynomial algebra, as flagged in Example (A Non-Archimedean Ordered Field)). An example with F Archimedean needs more algebra than this volume has, and we give none.

Marking scheme. (10 points) - 2: each χi fixes Q, with the reason (Proposition 6.1.1, or directly: induction for N, then negatives and quotients). - 3: links the rationals below x to those below χi(x), using that χi fixes Q and reflects order. Either gives the full equivalence ιF(q)<x⟺ιL(q)<χi(x), or, for the squeeze route of the next item, applies reflection by χ1 and χ2 to a rational strictly between χ1(x) and χ2(x). - 3: concludes χ1(x)=χ2(x) by rationals separating, or by a rational strictly between the two, invoking the Archimedean property of L. - 2: states the conclusion for every x∈F. Fatal errors (0 for the step): using density in L without a hypothesis that gives it; assuming χi surjective.

Exercise (C2: Dense image, and when the map is onto)

Let F be an Archimedean ordered field, L a complete ordered field, and φ:F→L the map of Theorem 6.2.3.

  • (a) Prove that φ(F) is dense in L: for all u<v in L there is x∈F with u<φ(x)<v.
  • (b) Prove that if F is complete, then φ is surjective.
  • (c) Prove that if φ is surjective, then F is complete.

Solution (Hint 1). For (b), given y∈L, look at the set S:=QL(y) and its copy ιF(S) in F.

Solution (Hint 2). For (b): ιF(S) is non-empty and bounded above in F; let x:=supFιF(S), and show QF(x)=S.

Solution. By Theorem 6.1.2, L is Archimedean.

  • (a) Let u<v in L. By Theorem 4.2.5 in L there is q∈Q with u<ιL(q)<v. By Step 3 of the proof of Theorem 6.2.3, ιL(q)=φ(ιF(q)), so x:=ιF(q) works.

  • (b) Let F be complete and y∈L. Let S:=QL(y). Since L is Archimedean, there are n,m∈N with −y<ιL(n) and y<ιL(m). So −n∈S, and every q∈S has ιL(q)<y<ιL(m), hence q<m and ιF(q)<ιF(m). So ιF(S) is non-empty and bounded above in F, and x:=supFιF(S) exists.

    We show QF(x)=S. (⊇) Let q∈S. By density in L there is r∈Q with ιL(q)<ιL(r)<y; then r∈S and q<r, so ιF(q)<ιF(r)≤x. (⊆) Let ιF(q)<x. Then ιF(q) is not an upper bound of ιF(S), so ιF(q)<ιF(r) for some r∈S; hence q<r and ιL(q)<ιL(r)<y, so q∈S.

    By (⋆), QL(φ(x))=QF(x)=S=QL(y), so φ(x)=y by Lemma 6.1.3 in L.

  • (c) Let φ be surjective. It is injective (strictly increasing), so bijective, with an inverse φ−1:L→F (Sets, Relations, and Cardinality, Theorem 1.5.4). Let A⊆F be non-empty with an upper bound b. Then φ(A) is non-empty and bounded above by φ(b), since a≤b gives φ(a)≤φ(b). As L is complete, s:=supφ(A) exists; let c:=φ−1(s). Upper bound: for a∈A, φ(a)≤s=φ(c), so a≤c, because φ reflects order (and is injective, for the case of equality). Least: if u is an upper bound of A, then φ(u) is an upper bound of φ(A), so φ(c)=s≤φ(u), and c≤u as before. So c=supA (Sets, Relations, and Cardinality, Definition 2.6.5), and F is complete.

So φ is an order-preserving field isomorphism exactly when F is complete. Section 6.3 reaches the “if” direction by a different route, which computes no supremum at all.

Marking scheme. (10 points) - 2: (a) L is Archimedean (Theorem 6.1.2), so some ιL(q) lies strictly between u and v, and ιL(q)=φ(ιF(q)) (Step 3). - 2: (b) ιF(S) is non-empty and bounded above in F (Archimedes in L, transferred through ιF), so x=supFιF(S) exists by completeness of F. - 2: (b) QF(x)=S, both inclusions; (⊇) uses density in L. - 1: (b) QL(φ(x))=QF(x)=QL(y) by (⋆), hence φ(x)=y by Lemma 6.1.3 in L. - 3: (c) for non-empty A⊆F bounded above by b, φ(A) is bounded above by φ(b), and φ−1(supφ(A)) is an upper bound of A and the least one, both using that φ reflects order and is bijective. Fatal errors (0 for the step): using density of ιL(Q) in L without the Archimedean property of L; in (b), taking a supremum in F before checking that the set is non-empty and bounded above; in (c), assuming φ−1 preserves order without justification.