S. D. V. Stephens

notes / Integrated Mathematics / chapter 17: The Real Numbers

17.4 Suprema, Completeness and the Archimedean Property

In progress: these notes stop partway.

The third group of axioms is a single axiom, completeness, which is about upper bounds of sets. This page defines the positive integers, sets up the language of upper bounds needed to state the axiom, and then derives its consequences.

Positive integers and the real line

Definition 17.4.1 (Inductive set). A set of real numbers is an inductive set if it has the following properties: the number is in the set, and for every in the set the number is also in the set.

is an inductive set, and so is .

Definition 17.4.2. A real number is called a positive integer if it belongs to every inductive set.

Real numbers can be pictured as points on a line:

A number line with arrows at both ends and tick marks labelled x, then dots, minus 1, 0, 1, dots, and y.

Upper bounds, maximum element and supremum

Suppose for every . Then is an upper bound of , and is bounded above. This also means that for every , is also an upper bound of . If moreover , then is the maximum element or largest member of . There can only be at most ONE such . We write this as . Thus: if and for all . A set with no upper bound is called unbounded above (example: the set of all real numbers, , has no ).

Example 17.4.3.

  1. : DNE (does not exist), and .
  2. : , and .
  3. : , but . (Does ?)

: is , while would be .

Remark (Correction). The set has no maximum element, so does not exist. The number is , and it is not in . The correct statement is that has a supremum but no maximum.

Two number lines. On the left, the closed interval S from 0 to 1 ends in a filled dot at 1 labelled max S, with the upper bounds hatched to its right. On the right, the interval T from 0 to 1 ends in an open dot at 1 labelled least upper bound or sup T, with the upper bounds hatched to its right.

The least upper bound is known as the supremum.

Theorem 17.4.4 (Theorem I.26). Two different numbers cannot both be least upper bounds for the same set.

Proof. Say we have two least upper bounds and . Since is a least upper bound, . Then since too is a least upper bound, we have . Thus and , so . ∎

We can now write, for a set , .

The least-upper-bound axiom

Definition 17.4.5 (Axiom 10). Every nonempty set of real numbers which is bounded above has a supremum; that is, there is a real number such that .

Moreover, does not need to be in : only when .

Similar to we have (the infimum), or the greatest lower bound, and is similar to : is the minimum element in .

Example 17.4.6. since, for example, Then and .

The Archimedean property

We must encounter consequences of the completeness axiom (Axiom 10).

Theorem 17.4.7 (Theorem I.28). The set of positive integers is unbounded above.

Proof. Assume is bounded above. Since , by Axiom 10 must have a least upper bound (), say . The number cannot be an upper bound of , as it is less than . So there must be at least one integer in such that . Thus . But since is in and , we arrive at a contradiction. ∎

Corollary 17.4.8 (Theorem I.29 / Corollary I.1). For every real , there must be a positive integer such that .

Proof. If there were no , then would be bounded by . However, by Theorem I.28 we know that is unbounded, so this cannot be. ∎

Corollary 17.4.9 (Theorem I.30 / Corollary I.2). If and if is an arbitrary real number, there exists a positive integer such that .

Proof. . Substitute for in Theorem I.29. ∎

Theorem I.30 is commonly known as the Archimedean property of the real number system. This, for the layman, means that any line, no matter how long, can be covered by a finite number of smaller lines. In other words, a small ruler, used enough times, can cover any given distance, no matter how large. However, there are “non-Archimedean geometries” in which this axiom is rejected. From this property we can formulate a theorem which will be useful in integral calculus.

Theorem 17.4.10 (Theorem I.31). If three real numbers , and satisfy the inequalities for every integer , then .

(I am confused about this... is an infinitesimal???)

Proof. If , Theorem I.30 tells us that there is a positive integer satisfying . This contradicts Theorem I.14. Hence we cannot have , so .

(↑ What?? How??) ∎

Remark (Correction). The “I.14” in the proof is the display number of the hypothesis ; Theorem I.14 is a different result. Indeed gives , which contradicts . Hence is impossible, and .

Fundamental properties of the supremum and infimum

Theorem 17.4.11 (Theorem I.32). Let be a given positive number and let be a set of real numbers.

(a) If has a supremum, then for some we have .

(b) If has an infimum, then for some we have .

(This establishes that the and can be changed by some sufficiently large .)

Remark (Correction). The theorem says that for every there is a point of within distance of (when the supremum exists), and likewise a point within of . The number is an arbitrary positive real number; it does not have to be large or to belong to , and small is the interesting case.

Proof. If we had for all , then would be an upper bound for smaller than its least upper bound. Therefore we must have for at least one . This proves (a). The proof for (b) is similar. ∎

Theorem 17.4.12 (Theorem I.33 (Additive property)). Given nonempty subsets and of , let denote the set (a) If each of and has a supremum, then has a supremum and .

(b) If each of and has an infimum, then has an infimum and .

Proof. Assume each of and has a supremum. If , then , where and . Therefore ; so is an upper bound for . This shows that has a supremum and that Now let be any positive integer. By Theorem I.32 (with ) there is an in and a in such that Adding these inequalities, we obtain since . Therefore we have shown that for every integer . By Theorem I.31, we must have . This proves (a), and the proof of (b) is similar. ∎

Theorem 17.4.13 (Theorem I.34). Given two nonempty subsets and of such that for every in and every in . Then has a supremum and has an infimum, and they satisfy .

Proof. Each in is an upper bound for . Therefore has a supremum which satisfies the inequality for all in . Hence is a lower bound for , so has an infimum which cannot be less than . In other words, we have , as asserted. ∎

Existence of square roots of nonnegative real numbers

  • , thus .
  • , , :
    • ; but is the least upper bound of , thus .
    • , which violates Axiom 10.
    • must be true.

Remark (Correction). The set is all of , which has no supremum, and the argument treats the number and the set as the same object. The intended result (Theorem I.35) is that every has a unique nonnegative square root; uniqueness follows from , and existence is proved as follows. Let ; it is nonempty () and bounded above by , so exists by Axiom 10. If , a slightly larger number is still in , contradicting that is an upper bound; if , a slightly smaller number is still an upper bound, contradicting that is the least one; so .

Exercises

(Sobbing and crying violently rn.)

Exercise 17.4.14. If and are arbitrary real numbers with , prove that there is at least one real satisfying .

Show solution

Let and be subsets of , and let and , where . Assume has a supremum of and has an infimum of . Thus . Thus exists. (Does this work?)

Sketch: , , with and : (???)

Remark (Correction). Assuming sets with already assumes that such a exists, and even then only follows. A proof: let . Adding and then to both sides of (Theorem I.18) gives , and multiplying by (Theorem I.19) gives .

Exercise 17.4.15. If , prove that there are integers and such that .

Show solution

We know there must exist some such that , by Theorem I.29. Assume is bounded below. Theorem I.27 tells us that there exists a greatest lower bound, say . Then , being more than , cannot be a lower bound for . Thus there is at least one integer such that . For this we have ; this cannot be, because is a greatest lower bound. Thus we have a contradiction. Now we state that there exists a such that is . We know this must be true because if not, there would be a greatest lower bound for , contradicting what we just proved. Thus replace with and with . We now have .

Remark (Correction). The second half does not produce an integer : it argues about real numbers below a real , which is a different matter. A proof: by Theorem I.29 there are positive integers and . Then is an integer with , so .

Exercise 17.4.16. If , prove that there is a positive integer such that .

Show solution

In the previous proof we created a corollary which proved that there exists an such that when is some number in . does not change this.

Remark (Correction). Having an integer does not by itself give . A proof: Theorem I.29 applied to the real number gives a positive integer , and multiplying by gives .

Exercise 17.4.17. If , prove that there is exactly one integer such that . This is called the greatest integer in and is denoted by . For example, , , .

(I’m honestly confused with what this is even saying.)

Exercise 17.4.18. If , prove that there is exactly one integer with .

Show solution

We know ; in terms of integers, is one more than , thus . Assume for integers. Then , but , thus this cannot be, and we must have one integer, where .

Remark (Correction). Taking is wrong for a general real : for the integer is , and can happen. A proof: by the previous exercise there is an integer with , that is, , so satisfies . If and both satisfy this, then , and two integers that close are equal.

Exercise 17.4.19. If with , prove that there exists at least one rational such that , and hence infinitely many.

Show solution

By an earlier exercise we know that there exist , where . Thus the inverse of must be true, thus , where there exist infinitely many .

Remark (Correction). The argument given does not prove the claim, and its should be (for real the claim is the first exercise in this section). A proof: by Theorem I.30 choose a positive integer with . Let , where is the greatest integer; then , so is rational with . Repeating the construction between and , and so on, gives infinitely many.

Exercise 17.4.20. If is rational, , and is irrational, prove that , , , and are all irrational.

Show solution

(a) Assume , where is irrational and is rational, is rational. Then . Then by this we know and , but that cannot be; thus results in an irrational number.

(b) . Follows from (a).

Remark (Correction). The step ‘so ’ needs a reason: is a difference of two rationals, hence rational. This contradicts being irrational.

Exercise 17.4.21. Is the sum of two irrational numbers always rational? Say that we add , where (irrational).