math.wiki
Analysis › The real numbers › Least upper bounds and completeness
Sign in

Least upper bounds and completeness

The rational numbers obey every law of an ordered field, yet no rational number squares to 2. What property of the real numbers guarantees that the numbers calculus needs are really there?

Where you are 4 in The real numbers

Why this exists

The real numbers as an ordered field derived algebra and inequalities from the field and order axioms. Every one of those axioms holds in \mathbb{Q}, yet x^2 = 2 has no rational solution (Irrational numbers). So those axioms alone cannot prove that \displaystyle \sqrt{2} exists, that a polynomial which changes sign has a zero, or that a bounded increasing sequence converges. On the The real number line, \mathbb{Q} has gaps.

The least-upper-bound property closes them: a nonempty set of reals with an upper bound has a least one. That one axiom, added to the ordered-field axioms, is what calculus rests on.

Definition

Let F be an ordered field and E \subseteq F.

  • b \in F is an upper bound of E if x \le b for every x \in E; then E is bounded above. Lower bound and bounded below are defined with b \le x, and E is bounded if it is bounded above and below.
  • s \in F is the least upper bound, or supremum, of E, written s = \sup E, if s is an upper bound of E and s \le b for every upper bound b of E.
  • t \in F is the greatest lower bound, or infimum, written t = \inf E, if t is a lower bound of E and b \le t for every lower bound b of E.
  • If \sup E exists and belongs to E, it is the maximum, \max E; if \inf E exists and belongs to E, it is the minimum, \min E.

F has the least-upper-bound property if every nonempty subset of F that is bounded above has a supremum in F.

The completeness axiom. \mathbb{R} is an ordered field with the least-upper-bound property. With The real numbers as an ordered field, this is everything the course assumes about the real numbers.

We write c + E = \{c + x : x \in E\}, cE = \{cx : x \in E\}, -E = (-1)E, A + B = \{a + b : a \in A, \ b \in B\} and AB = \{ab : a \in A, \ b \in B\}.

Intuition

Anything above an upper bound is again an upper bound, so the upper bounds of a set form a ray. The supremum is the left end of that ray, when it has one in F. It need not belong to the set: an open interval (Interval and set-builder notation) has its right endpoint as supremum and no largest element.

E
E
\sup E
\text{upper bounds of } E
Figure 1The upper bounds of E form a ray, and \sup E is its left end. Here \sup E is not an element of E.

In \mathbb{Q} the ray can lack a left end: every rational upper bound of \{x \in \mathbb{Q} : x^2 < 2\} has a smaller rational upper bound (Result 6).

Proofs use the second picture. An upper bound s is the supremum exactly when nothing smaller is an upper bound, that is, when every window just below s catches a point of E.

E
x
\varepsilon
s - \varepsilon
s = \sup E
Figure 2However small \varepsilon > 0 is, the window (s - \varepsilon, s] holds a point x of E. This is Result 2.

Results

Let F be an ordered field and E \subseteq F. Results of The real numbers as an ordered field are cited as OF1 to OF12.

1. Uniqueness. A set has at most one supremum and at most one infimum. An element of E that is an upper bound of E is \sup E; an element of E that is a lower bound is \inf E.

2. The approximation property. Let E be nonempty and s an upper bound of E. Then s = \sup E if and only if for every \varepsilon > 0 there is x \in E with s - \varepsilon < x \le s. Likewise, a lower bound t is \inf E if and only if for every \varepsilon > 0 there is x \in E with t \le x < t + \varepsilon.

3. Infimum from supremum. \inf E exists if and only if \sup(-E) does, and then \inf E = -\sup(-E). So if F has the least-upper-bound property, every nonempty subset that is bounded below has an infimum.

4. Comparison. Let A, B \subseteq \mathbb{R} be nonempty. If A \subseteq B and B is bounded above, then \sup A \le \sup B. If a \le b for every a \in A and b \in B, then \sup A and \inf B exist and \sup A \le \inf B.

5. Shifts and scalings. Let E \subseteq \mathbb{R} be nonempty and bounded above, and c \in \mathbb{R}. Then \sup(c + E) = c + \sup E, and \sup(cE) = c \sup E if c > 0. If instead E is bounded below and c < 0, then \sup(cE) = c \inf E. The same holds with \sup and \inf (and above and below) exchanged.

6. \mathbb{Q} is not complete. The set \{x \in \mathbb{Q} : x^2 < 2\} is nonempty and bounded above in \mathbb{Q} but has no supremum in \mathbb{Q}.

7. The square root of 2. There is exactly one positive real r with r^2 = 2, namely r = \sup\{x \in \mathbb{R} : x^2 < 2\}. It is irrational, so \mathbb{R} \setminus \mathbb{Q} is nonempty.

Why it is true

Result 1. Proof. Two suprema s, s' are each upper bounds and each at most every upper bound, so s \le s' and s' \le s, and s = s' by (O1); infima likewise. If m \in E is an upper bound and b is any upper bound, then m \le b because m \in E, so m = \sup E. \square

Result 2. Proof. Let s = \sup E and \varepsilon > 0. As s - \varepsilon < s (OF4), s - \varepsilon is not an upper bound, so some x \in E has x > s - \varepsilon, and x \le s. Conversely, suppose the window condition holds and an upper bound b has b < s. With \varepsilon = s - b, some x \in E has x > s - \varepsilon = b, which is impossible. So s \le b for every upper bound b (O1), and s = \sup E. The infimum half is Exercise 11. \square

Result 3. Proof. By OF4, b is a lower bound of E exactly when -b is an upper bound of -E. If s = \sup(-E), then -s is a lower bound of E, and every lower bound b has s \le -b, so b \le -s: thus \inf E = -s. The same steps with E and -E exchanged (note -(-E) = E) give the converse. If E is nonempty with lower bound b, then -E is nonempty with upper bound -b, so the least-upper-bound property gives \sup(-E), hence \inf E. \square

Result 4. Proof. \sup B is at least every element of B, hence of A, so \sup A \le \sup B. For the second claim, each b \in B is an upper bound of A, so \sup A exists and \sup A \le b. Then \sup A is a lower bound of B, so \inf B exists (Result 3) and \sup A \le \inf B. \square

Result 5. Proof. Let s = \sup E. For x \in E, c + x \le c + s, so c + s is an upper bound of c + E. If b is any upper bound of c + E, then b - c is an upper bound of E, so s \le b - c and c + s \le b. For c > 0, OF5 turns x \le s into cx \le cs; and if b is an upper bound of cE, then c^{-1}b is one of E (OF7), so s \le c^{-1}b and cs \le b. For c < 0, write cE = (-c)(-E) and use Result 3: \sup(cE) = (-c)\sup(-E) = (-c)(-\inf E) = c \inf E. The infimum statements follow by applying these to -E with Result 3. \square

Results 6 and 7. Both rest on a lemma valid in every ordered field F. Write A_F = \{x \in F : x^2 < 2\}.

Lemma. Let s \in F with s > 0.

  • (a) If s^2 < 2, then \displaystyle h = \frac{2 - s^2}{2s + 2} is positive and (s + h)^2 < 2.
  • (b) If s^2 > 2, then \displaystyle t = \frac{s^2 + 2}{2s} satisfies 0 < t < s, and t is an upper bound of A_F.

Proof. (a) h > 0 as a quotient of positives. Since s^2 + 2s > 0, 2 - s^2 < 2s + 2, so h < 1 and h^2 < h (OF5). Therefore

\begin{aligned} (s + h)^2 &= s^2 + 2sh + h^2 \\ &< s^2 + (2s + 2)h = 2. \end{aligned}

(b) t > 0, and \displaystyle s - t = \frac{s^2 - 2}{2s} > 0. Also

t^2 - 2 = \frac{(s^2 + 2)^2 - 8s^2}{4s^2} = \frac{(s^2 - 2)^2}{4s^2} > 0.

If some x \in A_F had x > t, then x > 0 and OF9 would give x^2 > t^2 > 2. So every element of A_F is at most t. \square

Both s + h and t are built from s by field operations, so they are rational when s is.

Proof of Result 6. 1 \in A_\mathbb{Q}, and 2 is an upper bound, since x > 2 gives x^2 > 4 (OF9). Suppose, for contradiction, that s = \sup A_\mathbb{Q} exists in \mathbb{Q}. Then s \ge 1, and s^2 \ne 2 since no rational number squares to 2. If s^2 < 2, the lemma puts s + h > s in A_\mathbb{Q}, so s is not an upper bound. If s^2 > 2, it gives a rational upper bound t < s, so s is not least. \square

Proof of Result 7. A_\mathbb{R} is nonempty and bounded above by 2, so r = \sup A_\mathbb{R} exists by the completeness axiom, and r \ge 1. The lemma with F = \mathbb{R} rules out r^2 < 2 and r^2 > 2 as above, so r^2 = 2. A second positive root equals r by OF9, and r is irrational by Irrational numbers. \square

Worked examples

Example 1

Sets cut out by inequalities

Find the supremum, infimum, maximum and minimum, where they exist, of E = \{x \in \mathbb{R} : x^2 - 4x < 5\} and of G = \{x \in \mathbb{R} : x^2 - 4x \le 5\} \cup \{7\}. Prove the claims about E with Result 2.

Solution

Since x^2 - 4x - 5 = (x - 5)(x + 1), the expression is negative exactly when the factors have opposite signs, that is, when -1 < x < 5. So E = (-1, 5) and G = [-1, 5] \cup \{7\}.

For E: 5 is an upper bound. Given \varepsilon > 0, let \displaystyle x = 5 - \frac{1}{2}\min\{\varepsilon, 6\}. Then 2 \le x < 5, so x \in E, and \displaystyle x \ge 5 - \frac{\varepsilon}{2} > 5 - \varepsilon. By Result 2, \sup E = 5. Likewise \displaystyle -1 + \frac{1}{2}\min\{\varepsilon, 6\} is a point of E below -1 + \varepsilon, so \inf E = -1. Neither is in E, so E has no maximum or minimum.

For G: 7 \in G is an upper bound, so \sup G = \max G = 7 by Result 1, and -1 \in G is a lower bound, so \inf G = \min G = -1.

Example 2

A set traced by a parameter

Let \displaystyle E = \left\{\frac{3t}{t + 1} : t > 0\right\}. Prove that \sup E = 3 and \inf E = 0, neither attained. For \displaystyle \varepsilon = \frac{1}{100}, name an element of E above 3 - \varepsilon.

Solution

Write \displaystyle \frac{3t}{t + 1} = 3 - \frac{3}{t + 1}. For t > 0, \displaystyle 0 < \frac{3}{t + 1} < 3, so every element of E lies strictly between 0 and 3.

Given \varepsilon > 0, take \displaystyle t = \frac{3}{\varepsilon}. Then

\frac{3}{t + 1} = \frac{3\varepsilon}{3 + \varepsilon} < \varepsilon,

so the element \displaystyle 3 - \frac{3}{t + 1} exceeds 3 - \varepsilon, and Result 2 gives \sup E = 3. With \displaystyle t = \frac{\varepsilon}{3} the element is \displaystyle \frac{3t}{t + 1} = \frac{3\varepsilon}{3 + \varepsilon} < \varepsilon, so \inf E = 0.

For \displaystyle \varepsilon = \frac{1}{100}, t = 300 gives \displaystyle \frac{900}{301}, which exceeds 2.99 because 2.99 \cdot 301 = 899.99. The t depends on \varepsilon, as the cutoff N will in Limits of sequences, epsilon and N.

Example 3

Laundry at the hostel

Amara and Samir run the laundry room at a hostel. A load costs \$4 plus \$1.50 per kilogram, so a load of w kilograms averages \displaystyle c(w) = 1.5 + \frac{4}{w} dollars per kilogram. (a) Find the infimum of \{c(w) : w > 0\}. Is it a minimum? (b) The washer holds at most 8 kilograms. Find the infimum of \{c(w) : 0 < w \le 8\}. Is it a minimum?

Solution

(a) Since \displaystyle \frac{4}{w} > 0, c(w) > 1.5: a lower bound no load reaches. Given \varepsilon > 0, the load \displaystyle w = \frac{8}{\varepsilon} has \displaystyle c(w) = 1.5 + \frac{\varepsilon}{2} < 1.5 + \varepsilon. By Result 2 the infimum is \$1.50 per kilogram, and there is no minimum.

(b) For 0 < w \le 8, OF7 gives \displaystyle \frac{1}{w} \ge \frac{1}{8}, so c(w) \ge 2, while c(8) = 2. A lower bound in the set is its infimum (Result 1), so the infimum is the minimum, \$2.00 per kilogram, at a full load. The cap on w turns a bare infimum into a minimum.

Example 4

The supremum of a sum

Prove that if A, B \subseteq \mathbb{R} are nonempty and bounded above, then \sup(A + B) = \sup A + \sup B. Then find the supremum and infimum of the set of all \displaystyle x + \frac{3t}{t + 1} with -1 < x < 5 and t > 0.

Solution

Proof. Let a = \sup A and b = \sup B. For x \in A and y \in B, x + y \le a + b by OF8, so a + b is an upper bound of A + B. Given \varepsilon > 0, Result 2 with \displaystyle \frac{\varepsilon}{2} gives x \in A with \displaystyle x > a - \frac{\varepsilon}{2} and y \in B with \displaystyle y > b - \frac{\varepsilon}{2}. Then x + y \in A + B and x + y > a + b - \varepsilon, so \sup(A + B) = a + b by Result 2. \square

Since -(A + B) = (-A) + (-B), Result 3 gives \inf(A + B) = \inf A + \inf B for sets bounded below. The set asked about is (-1, 5) + E with E from the second example: its supremum is 5 + 3 = 8 and its infimum is -1 + 0 = -1.

Example 5

Solving an equation with a supremum

Let g(x) = x^3 + x and E = \{x \in \mathbb{R} : g(x) < 3\}. Prove that r = \sup E exists and that r is the only real number with r^3 + r = 3.

Solution

Proof. g is strictly increasing: for x < y, g(y) - g(x) = (y - x)(y^2 + xy + x^2 + 1), and \displaystyle y^2 + xy + x^2 = \left(y + \frac{x}{2}\right)^2 + \frac{3x^2}{4} \ge 0.

Since g(1) = 2 < 3, 1 \in E; and x \ge 2 gives g(x) \ge g(2) = 10, so 2 is an upper bound. Thus r = \sup E exists and 1 \le r \le 2. For 0 < h \le 1,

\begin{aligned} g(r + h) - g(r) &= h(3r^2 + 3rh + h^2 + 1), \\ g(r) - g(r - h) &= h(3r^2 - 3rh + h^2 + 1), \end{aligned}

and each bracket is at most 3 \cdot 4 + 3 \cdot 2 + 1 + 1 = 20. Let D = g(r) - 3.

If D < 0, take \displaystyle h = \min\left\{1, -\frac{D}{40}\right\}. Then \displaystyle g(r + h) \le 3 + D + 20h \le 3 + \frac{D}{2} < 3, so r + h \in E, against r being an upper bound.

If D > 0, take \displaystyle h = \min\left\{1, \frac{D}{40}\right\}. Then \displaystyle g(r - h) \ge 3 + D - 20h \ge 3 + \frac{D}{2} > 3. Every x \in E has g(x) < g(r - h), so x < r - h as g is increasing: r - h is an upper bound below r, against r being least.

So D = 0, that is, r^3 + r = 3, and as g is strictly increasing no other real number works. \square

Numerically r \approx 1.2134. The same plan, with continuity in place of the explicit bound 20, proves the intermediate value theorem in Extreme and intermediate value theorems, proved.

Boundaries

The supremum need not be in the set. (-1, 5) has no maximum, and the unlimited hostel loads have no cheapest average.

Empty and unbounded sets have no supremum. Every real number is an upper bound of \varnothing, and an unbounded set has none. Some books set \sup \varnothing = -\infty and \sup E = \infty in the extended reals; this course does not.

The property belongs to the field. \{x \in \mathbb{Q} : x^2 < 2\} has a supremum in \mathbb{R} but not in \mathbb{Q}. That it equals \displaystyle \sqrt{2} uses the density of \mathbb{Q} from The Archimedean property and density of the rationals, which also supplies \displaystyle \sup\{1 - \frac{1}{n} : n \ge 1\} = 1. The sets here are traced by a real parameter, so the needed t or w can be written down.

Strict inequalities do not survive. In Result 4, a < b for every pair still gives only \sup A \le \inf B (Exercise 15).

Products need signs. \sup(AB) = \sup A \cdot \sup B for sets of positive numbers (Exercise 12) and fails for mixed signs (Exercise 8).

Where this goes next

The Archimedean property and density of the rationals uses completeness to show that the positive integers are not bounded above and that \mathbb{Q} is dense in \mathbb{R}. Countable and uncountable sets uses it, as nested intervals, to prove \mathbb{R} uncountable. After Limits of sequences, epsilon and N, The monotone convergence theorem shows a bounded increasing sequence converges to the supremum of its terms. The Riemann integral, defined defines the integral through suprema of lower sums and infima of upper sums, and Extreme and intermediate value theorems, proved proves the extreme and intermediate value theorems from suprema.

Exercises

Bounds of given sets

  1. Find the supremum, infimum, maximum and minimum, where they exist, of \{x \in \mathbb{R} : 2x^2 - 5x - 3 < 0\}.
  2. Do the same for \{x \in \mathbb{R} : \lvert 2x - 3 \rvert \le 5, \ x \ne 4\}.
  3. Do the same for \displaystyle \left\{\frac{2t + 5}{t + 1} : t \ge 0\right\}.
  4. Do the same for \{x^2 - 6x : 1 < x < 4\}.

The approximation property

  1. The set \displaystyle \left\{\frac{4t}{t + 3} : t > 0\right\} has supremum 4. Find every t > 0 whose element exceeds \displaystyle 4 - \frac{1}{10}, and give one such element as a fraction.
  2. Show that \{x \in \mathbb{R} : x^3 < 8\} has supremum 2, giving for each \varepsilon > 0 an element above 2 - \varepsilon, and that it has no lower bound.

Arithmetic with suprema

  1. Let A = (1, 3]. Use Result 5 to find the supremum and infimum of 5 - 2A = \{5 - 2a : a \in A\}, and say which is attained.
  2. Let A = [-2, 1] and B = [-3, 2]. Find \sup(AB) and \inf(AB), and compare \sup(AB) with \sup A \cdot \sup B.
  3. The hostel’s large washer charges \$7 per load plus \$1.20 per kilogram and holds at most 14 kilograms. Find the infimum of the average cost per kilogram over loads 0 < w \le 14, say whether it is a minimum, and decide whether the averages are bounded above.

Proofs

  1. Prove that every nonempty finite subset of an ordered field has a maximum and a minimum.
  2. Prove the infimum half of Result 2 directly from the definitions.
  3. Prove that if A, B \subseteq (0, \infty) are nonempty and bounded above, then \sup(AB) = \sup A \cdot \sup B.
  4. Prove that every positive real number a has exactly one positive square root.
  5. Let F be an ordered field in which every nonempty set that is bounded below has an infimum. Prove, without using -E, that F has the least-upper-bound property. (Consider the set of upper bounds.)
  6. Refute with a counterexample: “If a < b for every a \in A and b \in B, then \sup A < \inf B.”
  7. Prove that \{x \in \mathbb{Q} : x^2 < 2\} is bounded below in \mathbb{Q} but has no infimum in \mathbb{Q}.
Answers
  1. 2x^2 - 5x - 3 = (2x + 1)(x - 3) < 0 exactly when \displaystyle -\frac{1}{2} < x < 3. The set is \displaystyle \left(-\frac{1}{2}, 3\right), with supremum 3 and infimum \displaystyle -\frac{1}{2} as in the first example; neither is in the set, so there is no maximum or minimum.
  2. \lvert 2x - 3 \rvert \le 5 means -1 \le x \le 4; removing 4 leaves [-1, 4). The minimum and infimum are -1. The supremum is 4, since \displaystyle 4 - \frac{1}{2}\min\{\varepsilon, 5\} lies in the set above 4 - \varepsilon; there is no maximum.
  3. \displaystyle \frac{2t + 5}{t + 1} = 2 + \frac{3}{t + 1} with \displaystyle 0 < \frac{3}{t + 1} \le 3 for t \ge 0, so the set is (2, 5]: maximum and supremum 5 (at t = 0). With \displaystyle t = \frac{3}{\varepsilon}, \displaystyle \frac{3}{t + 1} < \varepsilon as in the second example, so the infimum is 2, not attained.
  4. x^2 - 6x = (x - 3)^2 - 9, and x - 3 \in (-2, 1) gives (x - 3)^2 \in [0, 4); each y \in [-9, -5) is attained at \displaystyle x = 3 - \sqrt{y + 9}. The set is [-9, -5): minimum and infimum -9, supremum -5, no maximum.
  5. \displaystyle \frac{4t}{t + 3} = 4 - \frac{12}{t + 3} > 4 - \frac{1}{10} exactly when t + 3 > 120, that is, t > 117. For example t = 120 gives \displaystyle \frac{480}{123} = \frac{160}{41}, and \displaystyle \frac{160}{41} > \frac{39}{10} since 1600 > 1599.
  6. For x < y, y^3 - x^3 = (y - x)(y^2 + xy + x^2) with \displaystyle y^2 + xy + x^2 = \left(y + \frac{x}{2}\right)^2 + \frac{3x^2}{4} > 0, as x, y are not both 0. So x^3 < 8 exactly when x < 2: the set is (-\infty, 2). The element \displaystyle 2 - \frac{\varepsilon}{2} lies above 2 - \varepsilon, so the supremum is 2. For any real b, \min\{b, 2\} - 1 is an element below b, so there is no lower bound.
  7. Result 5 with c = -2 gives \sup(-2A) = -2\inf A = -2 and \inf(-2A) = -2\sup A = -6, so \sup(5 - 2A) = 3 and \inf(5 - 2A) = -1. The infimum is attained at a = 3; the supremum would need a = 1 \notin A. So 5 - 2A = [-1, 3).
  8. Multiplying -2 \le a \le 1 by b (OF5) puts ab between -2b and b, and each of these lies between its values at b = -3 and b = 2. So every product lies between the corner products 6, -4, -3, 2, which are attained: \sup(AB) = \max(AB) = 6 and \inf(AB) = \min(AB) = -4. But \sup A \cdot \sup B = 1 \cdot 2 = 2 \ne 6.
  9. \displaystyle c(w) = 1.2 + \frac{7}{w} \ge 1.2 + \frac{1}{2} = 1.7 for 0 < w \le 14, with equality at w = 14: the minimum is \$1.70 per kilogram. Not bounded above: for M > 0, the load \displaystyle w = \min\{14, \frac{7}{M}\} has c(w) > M (it is 1.2 + M, or 1.7 when \displaystyle M \le \frac{1}{2}). A full large load is 30 cents per kilogram cheaper than a full small one.
  10. Proof. Induction on the number n \ge 1 of elements; \{x\} has maximum x. If every n-element set has a maximum and E has n + 1 elements, pick y \in E and let m be the maximum of E \setminus \{y\}. If y \le m, then m is an upper bound of E in E; otherwise y is. Result 1 gives \max E. Then \min E = -\max(-E) by Result 3. \square
  11. Proof. Let t be a lower bound of the nonempty set E. If t = \inf E and \varepsilon > 0, then t + \varepsilon is not a lower bound, so some x \in E has t \le x < t + \varepsilon. Conversely, if the window condition holds and a lower bound b has b > t, then \varepsilon = b - t gives x \in E with x < b, which is impossible. So b \le t for every lower bound b, and t = \inf E. \square
  12. Proof. Let a = \sup A > 0 and b = \sup B > 0. For x \in A, y \in B, xy \le ay \le ab (OF5), so ab is an upper bound of AB. Let u be any upper bound of AB and fix y \in B. Then \displaystyle x \le \frac{u}{y} for every x \in A (OF5, OF7), so \displaystyle a \le \frac{u}{y}, that is, \displaystyle y \le \frac{u}{a}. As y was arbitrary, \displaystyle b \le \frac{u}{a}, so ab \le u. \square
  13. Proof. Let P = \{x \in \mathbb{R} : x^2 < a\}. Then 0 \in P, and a + 1 is an upper bound since x > a + 1 gives x^2 > (a + 1)^2 > a (OF9); let r = \sup P. The element \displaystyle x_0 = \frac{a}{a + 1} has x_0^2 < x_0 < a, so r \ge x_0 > 0. If r^2 < a, then \displaystyle h = \frac{a - r^2}{2r + a + 1} has 0 < h < 1, and (r + h)^2 < r^2 + (2r + 1)h, which is below r^2 + (2r + a + 1)h = a; so r + h \in P, impossible. If r^2 > a, then \displaystyle t = \frac{r^2 + a}{2r} has \displaystyle r - t = \frac{r^2 - a}{2r} > 0 and \displaystyle t^2 - a = \frac{(r^2 - a)^2}{4r^2} > 0, so t < r is an upper bound of P (OF9), impossible. So r^2 = a, and a second positive root equals r by OF9. \square
  14. Proof. Let E be nonempty and bounded above, with set of upper bounds U \ne \varnothing. Every x \in E is a lower bound of U, so u_0 = \inf U exists, and x \le u_0 for each x \in E. So u_0 \in U, and u_0 \le u for every u \in U: u_0 = \sup E. \square
  15. Take A = (0, 1) and B = (1, 2). Every a < 1 < b, yet \sup A = 1 = \inf B.
  16. Proof. Let Q_2 = \{x \in \mathbb{Q} : x^2 < 2\}. It is bounded below by -2, since x < -2 gives x^2 > 4 (OF9 for -x > 2). As (-x)^2 = x^2, -Q_2 = Q_2. If Q_2 had an infimum t in \mathbb{Q}, Result 3, valid in every ordered field, would make -t the supremum of Q_2 in \mathbb{Q}, against Result 6. \square
math.wiki · updated 2026-11-17 · CC-BY-SA-4.0