The real numbers as an ordered field
Calculus leaned on rules such as a negative times a negative is positive, and multiplying an inequality by a negative number reverses it. Where do those rules come from, and why do they hold for the rational numbers as well as the real ones?
Where you are 4 in The real numbers
Why this exists
Calculus flipped inequalities multiplied by -2 and bounded \lvert a + b \rvert by \lvert a \rvert + \lvert b \rvert without proof. Real analysis proves calculus, so it must say exactly what it assumes about \mathbb{R}.
The answer is short: \mathbb{R} is an ordered field with the least-upper-bound property. This lesson takes the first half. It lists the field and order axioms and derives from them the rules of algebra and of inequalities that later proofs use, ending with the absolute value and the triangle inequality. The second half is Least upper bounds and completeness.
Throughout the course, \mathbb{N} = \{0, 1, 2, \dots\}, as in Countable sets. Some books, Lebl’s Basic Analysis among them, start \mathbb{N} at 1; where such a book writes n \in \mathbb{N}, read “every positive integer n”, that is, n \ge 1.
Definition
A field is a set F with two operations, addition x + y and multiplication xy, such that for all x, y, z \in F:
| Addition | Multiplication | |
|---|---|---|
| closure | (A1) x + y \in F | (M1) xy \in F |
| commutative | (A2) x + y = y + x | (M2) xy = yx |
| associative | (A3) (x + y) + z = x + (y + z) | (M3) (xy)z = x(yz) |
| identity | (A4) some 0 \in F has x + 0 = x for every x | (M4) some 1 \in F with 1 \ne 0 has x1 = x for every x |
| inverse | (A5) each x has some -x \in F with x + (-x) = 0 | (M5) each x \ne 0 has some x^{-1} \in F with xx^{-1} = 1 |
and (D), the distributive law, x(y + z) = xy + xz. By (A2) and (M2) each law also holds with the sides of a sum or product swapped, for example 0 + x = x and (y + z)x = yx + zx.
Subtraction and division are abbreviations: x - y means x + (-y), and for y \ne 0, x/y or \displaystyle \frac{x}{y} means xy^{-1}. Also x^1 = x, x^{n+1} = x^n x, 2 = 1 + 1, and for a positive integer n, n \cdot 1 is the sum of n ones.
An ordered field is a field F with a relation < such that for all x, y, z \in F:
- (O1) trichotomy: exactly one of x < y, x = y, y < x holds;
- (O2) transitivity: if x < y and y < z, then x < z;
- (O3) if x < y, then x + z < y + z;
- (O4) if 0 < x and 0 < y, then 0 < xy.
We write y > x for x < y, and x \le y for “x < y or x = y”. The element x is positive if x > 0, negative if x < 0, nonnegative if x \ge 0. The absolute value of x is
The real numbers. We assume there is an ordered field \mathbb{R} that also has the least-upper-bound property. Building it from \mathbb{Q}, and showing it is unique up to renaming its elements, is beyond this course. Every result below uses only (A1) to (O4), so it holds in \mathbb{R}, in \mathbb{Q}, and in every other ordered field.
Intuition
Commutative, associative and distributive properties and Identity and inverse elements stated these laws as facts about numbers. Here they are the definition, and everything else is a theorem. That (-1)(-1) = 1, or that 1 > 0, is forced by the axioms, just as Proving statements about divisibility turned “a \mid b” into “b = ar” and worked from there. On the number line, (O3) says sliding the line keeps every pair in order; multiplying by -1 reflects it, which is why it reverses inequalities.
Results
Let F be a field, and in Results 4 to 11 an ordered field; x, y, z, w \in F.
1. Uniqueness and cancellation. The elements 0 and 1 are unique, each x has exactly one additive inverse, and each x \ne 0 exactly one multiplicative inverse. If x + z = y + z, then x = y. If xz = yz and z \ne 0, then x = y.
2. Signs. 0x = 0, (-1)x = -x, -(-x) = x, (-x)y = -(xy), and (-x)(-y) = xy.
3. No zero divisors. If xy = 0, then x = 0 or y = 0.
4. Positive and negative. x > 0 if and only if -x < 0. Also x < y if and only if y - x > 0, and if and only if -y < -x.
5. Multiplying an inequality. Let x < y. If z > 0, then xz < yz; if z < 0, then xz > yz. If x \le y and z \ge 0, then xz \le yz.
6. Squares. If x \ne 0, then x^2 > 0; so x^2 \ge 0 for every x, 1 > 0 and -1 < 0.
7. Reciprocals. If x > 0, then x^{-1} > 0. If 0 < x < y, then 0 < y^{-1} < x^{-1}.
8. Adding inequalities. If x < y and z \le w, then x + z < y + w; if x \le y and z \le w, then x + z \le y + w.
9. Squares of nonnegatives. If x \ge 0 and y \ge 0, then x < y if and only if x^2 < y^2, and x = y if and only if x^2 = y^2.
10. Absolute value.
- (a) \lvert x \rvert \ge 0, and \lvert x \rvert = 0 if and only if x = 0.
- (b) \lvert -x \rvert = \lvert x \rvert, so \lvert x - y \rvert = \lvert y - x \rvert.
- (c) \lvert x \rvert^2 = x^2 and \lvert xy \rvert = \lvert x \rvert \lvert y \rvert.
- (d) \lvert x \rvert \le y if and only if -y \le x \le y; and \lvert x \rvert < y if and only if -y < x < y.
- (e) -\lvert x \rvert \le x \le \lvert x \rvert.
11. Triangle inequalities. For all x, y \in F,
12. The number systems. \mathbb{Q}, with x < y meaning \displaystyle y - x = \frac{m}{n} for some positive integers m and n, is an ordered field. \mathbb{R} is one by assumption. \mathbb{Z} is not a field.
Why it is true
Result 1. Proof. If 0 and 0' both satisfy (A4), then 0 = 0 + 0' = 0', using 0' and then 0 as the identity; likewise for 1. If x + y = 0 and x + y' = 0, then by (A4), (A3), (A2) and (A4),
Products work the same way. If x + z = y + z, adding -z gives x = y by (A3), (A5), (A4); if xz = yz with z \ne 0, multiply by z^{-1}. \square
Result 2. Proof. By (D), 0x + 0x = (0 + 0)x = 0x = 0x + 0, so 0x = 0 by cancellation. Next, x + (-1)x = (1 + (-1))x = 0x = 0, so (-1)x is an additive inverse of x, hence -x by uniqueness. Since x + (-x) = 0, x is an additive inverse of -x: -(-x) = x. Then (-x)y = ((-1)x)y = (-1)(xy) = -(xy) by (M3), and using this twice with (M2), (-x)(-y) = -(-(xy)) = xy. \square
Result 3. Proof. If xy = 0 and x \ne 0, then y = (x^{-1}x)y = x^{-1}(xy) = x^{-1}0 = 0 by (M4), (M5), (M3) and Result 2. \square
Result 4. Proof. If x > 0, (O3) with z = -x gives 0 + (-x) < x + (-x), that is, -x < 0; if -x < 0, adding x gives 0 < x. Adding -x, and adding x, turn x < y into 0 < y - x and back. Finally (-x) - (-y) = y - x by Result 2 and (A2), so -y < -x exactly when y - x > 0. \square
Result 5. Proof. Let x < y and z > 0. Then y - x > 0 (Result 4), so z(y - x) = zy - zx > 0 by (O4), (D) and Result 2, and Result 4 gives zx < zy. If z < 0, then -z > 0, so (-z)x < (-z)y, that is, -(xz) < -(yz), and Result 4 gives yz < xz. In the last claim, if z = 0 or x = y then xz = yz; otherwise the first case applies. \square
Result 6. Proof. By (O1), x > 0 or x < 0. If x > 0, (O4) gives xx > 0. If x < 0, then -x > 0 and x^2 = (-x)(-x) > 0 by Result 2 and (O4). Since 1 \ne 0, 1 = 1^2 > 0, and -1 < 0 by Result 4. \square
Result 7. Proof. Let x > 0. If x^{-1} = 0, then 1 = xx^{-1} = 0, against (M4). If x^{-1} < 0, Result 5 applied to 0 < x gives 0 \cdot x^{-1} > xx^{-1}, that is, 0 > 1, against Result 6. So x^{-1} > 0. If 0 < x < y, then x^{-1}y^{-1} > 0 by (O4), and multiplying x < y by it (Result 5) gives y^{-1} < x^{-1}. \square
Result 8. Proof. From x < y, (O3) gives x + z < y + z. If z = w we are done; if z < w, (O3) gives y + z < y + w, and (O2) finishes. The version with \le is the same argument allowing equalities. \square
Result 9. Proof. Let x, y \ge 0 and x < y, so y > 0. If x = 0, then x^2 = 0 < y^2 by Result 6; if x > 0, then x^2 < xy < y^2 by Result 5 twice. Conversely, if x < y fails, then y \le x, so y^2 \le x^2 by what was just shown, and x^2 < y^2 fails. If x \ne y, one is smaller, so the squares differ. \square
Result 10. Proof. (a) If x \ge 0, \lvert x \rvert = x \ge 0, which is 0 only for x = 0. If x < 0, \lvert x \rvert = -x > 0 by Result 4.
(b) If x > 0, then \lvert -x \rvert = -(-x) = x; if x < 0, then \lvert -x \rvert = -x; and -0 = 0. Also y - x = -(x - y), as their sum is 0.
(c) \lvert x \rvert is x or -x, and (-x)(-x) = x^2, so \lvert x \rvert^2 = x^2. Then \lvert xy \rvert^2 = x^2 y^2 = (\lvert x \rvert \lvert y \rvert)^2, both sides’ bases are nonnegative (part (a), Result 5), and Result 9 makes them equal.
(d) First, x \le \lvert x \rvert and -x \le \lvert x \rvert: if x \ge 0, then -x \le 0 \le x = \lvert x \rvert; if x < 0, then x < 0 < -x = \lvert x \rvert. So if \lvert x \rvert \le y, then x \le y and -x \le y, which gives -y \le x (Result 4). Conversely, if -y \le x \le y, then -x \le y, and \lvert x \rvert is x or -x, so \lvert x \rvert \le y. The strict version has the same proof.
(e) is (d) with y = \lvert x \rvert. \square
Result 11. Proof. By Result 10(e), -\lvert x \rvert \le x \le \lvert x \rvert and -\lvert y \rvert \le y \le \lvert y \rvert. Adding (Result 8), -(\lvert x \rvert + \lvert y \rvert) \le x + y \le \lvert x \rvert + \lvert y \rvert, and Result 10(d) gives the first inequality. For the second, write x = (x - y) + y: \lvert x \rvert \le \lvert x - y \rvert + \lvert y \rvert, so \lvert x \rvert - \lvert y \rvert \le \lvert x - y \rvert. Swapping x and y and using Result 10(b), \lvert y \rvert - \lvert x \rvert \le \lvert x - y \rvert. Result 10(d) finishes. \square
Result 12. The field axioms for \mathbb{Q} are the rules of fraction arithmetic (Rational numbers), which follow from those of \mathbb{Z}; the order axioms are proved in the fourth worked example. \mathbb{Z} fails (M5): 2m is even for every integer m, so 2m \ne 1. \square
Worked examples
Difference of squares from the axioms
Prove that (x + y)(x - y) = x^2 - y^2 in every field, naming the reason for each step.
SolutionProof.
| Statement | Reason |
|---|---|
| (x + y)(x - y) = (x + y)x + (x + y)(-y) | definition of x - y; (D) |
| = (x^2 + yx) + (x(-y) + y(-y)) | (D) with (M2) |
| = (x^2 + xy) + (-(xy) + (-(y^2))) | Result 2; (M2) |
| = x^2 + ((xy + (-(xy))) + (-(y^2))) | (A3) twice |
| = x^2 + (0 + (-(y^2))) = x^2 - y^2 | (A5), (A4) |
\square
The step yx = xy in the third line cannot be dropped: for square matrices, where multiplication is not commutative, (A + B)(A - B) = A^2 - B^2 + BA - AB, and the identity can fail.
Midpoints
Prove that in every ordered field 2 > 0, and that if a < b then \displaystyle a < \frac{a + b}{2} < b. Find the midpoint of \displaystyle \frac{3}{7} and \displaystyle \frac{4}{9} and check the claim.
SolutionProof. By Result 6, 1 > 0, so (O3) gives 1 + 1 > 0 + 1 = 1, and (O2) gives 2 > 0; so 2^{-1} exists and is positive (Result 7). From a < b, (O3) gives a + a < a + b < b + b. By (D), a + a = (1 + 1)a = 2a and b + b = 2b. Multiplying 2a < a + b < 2b by 2^{-1} > 0 (Result 5) gives a < (a + b)2^{-1} < b. \square
In \mathbb{Q}, \displaystyle \frac{3}{7} = \frac{54}{126} and \displaystyle \frac{4}{9} = \frac{56}{126}, so the midpoint is \displaystyle \frac{55}{126}, and \displaystyle \frac{54}{126} < \frac{55}{126} < \frac{56}{126}.
A bound with the triangle inequality
Prove that if \lvert x - 3 \rvert < 1, then \lvert x^2 - 9 \rvert \le 7\lvert x - 3 \rvert. Deduce a bound for \lvert x^2 - 9 \rvert when \displaystyle \lvert x - 3 \rvert < \frac{1}{50}.
SolutionBy Result 10(c), \lvert x^2 - 9 \rvert = \lvert x - 3 \rvert \lvert x + 3 \rvert. Write the second factor around the point 3: \lvert x + 3 \rvert = \lvert (x - 3) + 6 \rvert, which by the triangle inequality and Result 8 is at most \lvert x - 3 \rvert + 6 < 1 + 6 = 7. Multiplying by \lvert x - 3 \rvert \ge 0 (Result 5), \lvert x^2 - 9 \rvert \le 7\lvert x - 3 \rvert.
If \displaystyle \lvert x - 3 \rvert < \frac{1}{50}, then also \lvert x - 3 \rvert < 1, so \displaystyle \lvert x^2 - 9 \rvert < \frac{7}{50}.
No constant below 7 would do: on (2, 4) the values of x + 3 fill (5, 7).
The rationals are ordered
Assuming the familiar facts about \mathbb{Z} (each integer is positive, zero or negative, exactly one; sums and products of positive integers are positive), prove that the order of Result 12 satisfies (O1) to (O4).
SolutionProof. Call a rational positive if it is \displaystyle \frac{m}{n} with m, n positive integers. Sums and products of positives are positive: \displaystyle \frac{m}{n} + \frac{p}{q} = \frac{mq + pn}{nq} and \displaystyle \frac{m}{n} \cdot \frac{p}{q} = \frac{mp}{nq}.
(O1). Write \displaystyle y - x = \frac{a}{b} with b \ge 1. If a > 0, then x < y; if a = 0, then x = y; if a < 0, then \displaystyle x - y = \frac{-a}{b} is positive and y < x. At most one holds: a positive \displaystyle \frac{m}{n} has m \ne 0, so 0 is not positive. That rules out x = y with either inequality, and x < y with y < x would make (y - x) + (x - y) = 0 positive.
(O2). z - x = (z - y) + (y - x) is a sum of positives. (O3). (y + z) - (x + z) = y - x. (O4). A product of positives is positive. \square
Smaller than every positive tolerance
Prove that if x \le y + \varepsilon for every \varepsilon > 0, then x \le y.
SolutionProof. By contradiction: suppose x > y. Then x - y > 0 (Result 4), so \displaystyle \varepsilon = \frac{x - y}{2} > 0 (Results 7 and 5). For this \varepsilon, \displaystyle y + \varepsilon = \frac{x + y}{2}, which is less than x by the midpoint example, since y < x. That contradicts x \le y + \varepsilon. By (O1), x \le y. \square
Yuna and Kofi’s orrery
Yuna and Kofi are building an orrery. One moon’s arm clicks around five marks, 0 to 4, and its arithmetic is done with remainders on division by 5: this is \mathbb{Z}_5 (Modular arithmetic). Every field axiom except (M5) holds in \mathbb{Z}_n for each n \ge 2, because remainders respect sums and products of integers. (a) Show that (M5) holds in \mathbb{Z}_5. (b) Solve 2x + 3 = 1 there. (c) Prove that no relation < makes \mathbb{Z}_5 an ordered field, and that every ordered field is infinite.
Solution(a) 1 \cdot 1 = 1, 2 \cdot 3 = 6 = 1 and 4 \cdot 4 = 16 = 1 in \mathbb{Z}_5, so 1^{-1} = 1, 2^{-1} = 3, 3^{-1} = 2 and 4^{-1} = 4.
(b) Add -3 = 2 to both sides: 2x = 1 + 2 = 3. Multiply by 2^{-1} = 3: x = 9 = 4. Check: 2 \cdot 4 + 3 = 11 = 1. By Result 1 the solution is unique.
(c) Proof. In every ordered field, n \cdot 1 > 0 for every positive integer n, by induction: 1 > 0 is Result 6, and if n \cdot 1 > 0, then (O3) and (O2) give (n + 1) \cdot 1 = n \cdot 1 + 1 > 0 + 1 > 0. In \mathbb{Z}_5, 5 \cdot 1 = 0, so an order would give 0 > 0, against (O1). In any ordered field, if m > n then m \cdot 1 = n \cdot 1 + (m - n) \cdot 1 > n \cdot 1 by (A3) and (O3), so 1, 2 \cdot 1, 3 \cdot 1, \dots are all different. \square
Boundaries
An ordered field is not yet \mathbb{R}. \mathbb{Q} satisfies every axiom here, yet x^2 = 2 has no solution in it (Irrational numbers). So nothing in this lesson can prove that \displaystyle \sqrt{2} exists; that takes the least-upper-bound property.
Fields that cannot be ordered. Finite fields such as \mathbb{Z}_5 cannot (the orrery example), and neither can \mathbb{C} (Exercise 12). \mathbb{Z} obeys (O1) to (O4) but is not a field.
Inequalities need their hypotheses. Products need positivity: -3 < 1, yet (-3)(-3) > 1 \cdot 1. Reciprocals reverse only within one sign: -1 < 2 and (-1)^{-1} < 2^{-1}.
Where this goes next
Least upper bounds and completeness adds the least-upper-bound property, which separates \mathbb{R} from \mathbb{Q}; The Archimedean property and density of the rationals follows from it. The triangle inequality drives every estimate in Limits of sequences, epsilon and N.
Exercises
Computing in fields
- In \mathbb{Z}_7, find the multiplicative inverse of each of 1, 2, 3, 4, 5, 6.
- In \mathbb{Z}_7, solve 3x + 5 = 2.
- Find every solution of x^2 = 1 in \mathbb{Z}_7 and in \mathbb{Z}_8. Explain the difference.
Inequalities and bounds
- Find every real x with \lvert 3x + 1 \rvert \le 2\lvert x \rvert, using Result 9.
- Find every real x with \lvert x - 1 \rvert + \lvert x + 2 \rvert < 5.
- Use the triangle inequality to find M with \lvert x^3 - 2x^2 + 7 \rvert \le M for -1 \le x \le 2. Then find the smallest such M.
Proofs from the field axioms
- Prove that -(x + y) = (-x) + (-y) in every field.
- Prove that if x \ne 0 and y \ne 0, then xy \ne 0 and (xy)^{-1} = x^{-1}y^{-1}.
Proofs from the order axioms
- Prove that if 0 \le x < y, then x^n < y^n for every positive integer n.
- Prove that x^2 + y^2 \ge 2xy, with equality only when x = y. Deduce that \displaystyle \frac{x}{y} + \frac{y}{x} \ge 2 when x, y > 0.
- Prove that if \lvert x \rvert < \varepsilon for every \varepsilon > 0, then x = 0.
- Prove that \mathbb{C}, a field containing i with i^2 = -1, is not an ordered field under any relation <.
Proofs about absolute value
- Prove that \lvert x + y \rvert = \lvert x \rvert + \lvert y \rvert if and only if xy \ge 0.
- Prove that \lvert x_1 + \dots + x_n \rvert \le \lvert x_1 \rvert + \dots + \lvert x_n \rvert for every positive integer n.
- Prove that if \displaystyle \lvert x - a \rvert < \frac{\varepsilon}{2} and \displaystyle \lvert y - a \rvert < \frac{\varepsilon}{2}, then \lvert x - y \rvert < \varepsilon. Deduce that if a \ne b and \displaystyle \varepsilon = \frac{\lvert a - b \rvert}{2}, no x has both \lvert x - a \rvert < \varepsilon and \lvert x - b \rvert < \varepsilon.
Answers
- 1^{-1} = 1, 2^{-1} = 4, 3^{-1} = 5, 4^{-1} = 2, 5^{-1} = 3, 6^{-1} = 6, since 8, 15 and 36 all leave remainder 1 on division by 7.
- 3x = 2 - 5 = -3 = 4, so x = 3^{-1} \cdot 4 = 5 \cdot 4 = 20 = 6. Check: 3 \cdot 6 + 5 = 23 = 2.
- In \mathbb{Z}_7: x = 1 and x = 6. In \mathbb{Z}_8: x = 1, 3, 5, 7. In a field, (x - 1)(x + 1) = 0 forces x = 1 or x = -1 by Result 3. \mathbb{Z}_8 is not a field: (3 - 1)(3 + 1) = 8 = 0 there, with neither factor 0.
- Both sides are nonnegative, so by Result 9 the inequality holds exactly when (3x + 1)^2 \le 4x^2, that is, 5x^2 + 6x + 1 \le 0, or (5x + 1)(x + 1) \le 0. The factors have opposite signs or one is 0: \displaystyle -1 \le x \le -\frac{1}{5}.
- For x \ge 1 the sum is 2x + 1 < 5, so x < 2. For -2 \le x < 1 it is 3 < 5, always true. For x < -2 it is -2x - 1 < 5, so x > -3. Together: -3 < x < 2.
- The triangle inequality gives at most \lvert x \rvert^3 + 2\lvert x \rvert^2 + 7, and \lvert x \rvert \le 2 makes that at most 8 + 8 + 7 = 23. The values at the endpoints and critical points -1, 0, \displaystyle \frac{4}{3}, 2 are 4, 7, \displaystyle \frac{157}{27}, 7, all positive, so the smallest M is 7.
- Proof. (x + y) + ((-x) + (-y)) = (x + (-x)) + (y + (-y)) = 0 by (A2), (A3), (A5) and (A4), so (-x) + (-y) is an additive inverse of x + y, and Result 1 makes it -(x + y). \square
- Proof. If xy = 0, Result 3 gives x = 0 or y = 0; so xy \ne 0. By (M2) and (M3), (xy)(x^{-1}y^{-1}) = (xx^{-1})(yy^{-1}) = 1, and the inverse of xy is unique (Result 1). \square
- Proof. Induction on n; n = 1 is the hypothesis. Suppose x^n < y^n. Since x \ge 0, x^n \ge 0 (Result 5, repeatedly), so x^{n+1} = x^n x \le x^n y by Result 5. Since y > 0, x^n y < y^n y = y^{n+1}. By (O2), x^{n+1} < y^{n+1}. \square
- Proof. x^2 + y^2 - 2xy = (x - y)^2 \ge 0 by Result 6, and it is 0 only when x - y = 0, since z^2 > 0 for z \ne 0. For x, y > 0, (xy)^{-1} > 0 (O4 and Result 7), and \displaystyle \frac{x}{y} + \frac{y}{x} - 2 equals (x - y)^2 (xy)^{-1}, which is \ge 0 by Result 5. \square
- Proof. If x \ne 0, then \lvert x \rvert > 0 (Result 10(a)); with \varepsilon = \lvert x \rvert the hypothesis gives \lvert x \rvert < \lvert x \rvert, against (O1). \square
- Proof. Suppose < made \mathbb{C} an ordered field. Since i \ne 0, Result 6 gives i^2 > 0, that is, -1 > 0. Result 6 also gives -1 < 0, against (O1). \square
- Proof. Both sides are nonnegative, so by Result 9 they are equal exactly when their squares are. By Result 10(c), \lvert x + y \rvert^2 = x^2 + 2xy + y^2 and (\lvert x \rvert + \lvert y \rvert)^2 = x^2 + 2\lvert xy \rvert + y^2. These agree exactly when xy = \lvert xy \rvert (cancel, then multiply by 2^{-1}), that is, when xy \ge 0. \square
- Proof. Induction on n; n = 1 is an equality. Suppose it holds for n and let s = x_1 + \dots + x_n. The triangle inequality gives \lvert s + x_{n+1} \rvert \le \lvert s \rvert + \lvert x_{n+1} \rvert, and the hypothesis bounds \lvert s \rvert by \lvert x_1 \rvert + \dots + \lvert x_n \rvert; add \lvert x_{n+1} \rvert (Result 8). \square
- Proof. Write x - y = (x - a) + (a - y). By Results 11 and 10(b), \lvert x - y \rvert \le \lvert x - a \rvert + \lvert y - a \rvert, and by Result 8 this is less than \displaystyle \frac{\varepsilon}{2} + \frac{\varepsilon}{2} = \varepsilon. If some x had \lvert x - a \rvert < \varepsilon and \lvert x - b \rvert < \varepsilon with \displaystyle \varepsilon = \frac{\lvert a - b \rvert}{2}, the same steps would give \lvert a - b \rvert < 2\varepsilon = \lvert a - b \rvert, against (O1). \square