Chapter I · Real variables

Real numbers

§8 25 min

Sections of all the rationals, not only the positive ones, give the real numbers, with zero, positive and negative numbers distinguished by where 0 falls.

Why this matters

The provisional definition of the last lesson only produced positive numbers, because it cut the positive rationals. Real numbers must also include zero and the negatives, and they must include the old rationals in some recognisable form. In §8 Hardy makes the final definition: cut all the rationals. With one simple test (where does 00 go?) every real number is then exactly one of positive, zero or negative.

The idea

Nothing in the idea of a section needs the rationals to be positive. Any condition that splits all the rationals into a lower class and an upper class, with everything in the lower class below everything in the upper class, does the same job. For example:

  • x+1<0x + 1 < 0 (lower) against x+1≥0x + 1 \ge 0 (upper) splits the rationals at −1-1, with −1-1 itself the least member of the upper class;
  • x3<−2x^3 < -2 against x3>−2x^3 > -2 splits them at a place where no rational sits, which we shall call −23-\sqrt[3]{2};
  • x<0x < 0 against x≥0x \ge 0 splits them at 00.

Once negative rationals are included, the position of 00 tells us the sign of the section. If 00 lies in the lower class, the number is to the right of 00; if some negative rational lies in the upper class, it is to the left; if 00 is the least member of the upper class, the number is 00.

Explore Where does 00 fall?

Try this. Choose each section in turn. For each one, use the test fraction to ask whether 00 (that is, p=0p = 0) is in LL or in RR, and then test a negative fraction such as −1/2-1/2.

Section
L={x:x3<−2}L=\{x : x^3 < -2\}, R = all other rationals
LR−1.450−1.400−1.350−1.300−1.250−1.200−1.150−1.100−1.050−1.3−1.2
  • L (lower class)
  • R (upper class)
  • gap not yet sorted
  • your fraction
With 1 decimal place: −1.3 is in L and −1.2 is in R; the gap between them has width 0.1. Your fraction 1712\tfrac{17}{12} ≈ 1.4166667 is in R.
Neither class has an end member: whatever rational you name in L, a larger one is still in L, and whatever you name in R, a smaller one is still in R. The cut defines a new number, α, sitting in the gap.
The upper class contains negative rationals (for example the decimal shown in R), so this section is a negative real number. Neither class has an end member: it is a negative irrational number.
What to notice

For 2\sqrt2 and 1/21/2 the rational 00 is in LL: positive numbers. For −1-1 and −23-\sqrt[3]{2} some negative rationals are already in RR: negative numbers. For the section labelled 00, all negatives are in LL and 00 is the least member of RR. The rational sections (−1-1, 00, 1/21/2) have a least member of RR; the irrational ones have no end member at all.

Precise statements

Throughout, a section of the rationals means a division of all the rational numbers into two classes LL and RR, both non-empty, with every member of LL less than every member of RR. As in §7, there are two sections at each rational aa, and we keep the one in which aa belongs to the upper class.

Definition Real number Hardy: Definitions, §8

A section of the rational numbers in which both classes exist and the lower class has no greatest member is called a real number, or simply a number. A real number that does not correspond to a rational number (that is, whose upper class has no least member) is called irrational. A real number that does correspond to a rational number aa is called rational too, and denoted aa.

Definition Positive, negative, zero Hardy: §8

The real number zero is the section whose lower class is exactly the negative rationals. A real number is positive if its lower class contains some positive rational, and negative if its upper class contains some negative rational.

Theorem Every real number has exactly one sign

Every real number is exactly one of: positive, zero, negative.

Theorem The positive real numbers are the old ones Hardy: §8

Adding 00 and all negative rationals to the lower class turns a positive real number in the sense of §7 into a positive real number in the sense of §8, and every positive real number arises in this way from exactly one section of the positive rationals.

Proofs

Proof Every real number has exactly one sign
Strategy

Split into cases by the class containing the rational 00, and use two facts: the lower class lies entirely below the upper class, and the lower class has no greatest member.

Case 0∈L0 \in L: the number is positive and not negative

LL has no greatest member, so some l∈Ll \in L exceeds 00: a positive rational in LL, so the number is positive. No negative rational rr can be in RR, because then the member 00 of LL would exceed the member rr of RR.

Case 0∈R0 \in R: the number is not positive

A positive rational ll in LL would exceed the member 00 of RR, which is impossible.

Within the case 0∈R0 \in R, either zero or negative

If every negative rational is in LL, then LL is exactly the set of negative rationals (positives and 00 are excluded by the previous step), and the number is zero. Otherwise some negative rational is in RR, and the number is negative. These two cannot both happen, by their definitions.

Conclude

The two cases 0∈L0 \in L and 0∈R0 \in R exhaust all possibilities and exclude each other, so exactly one of the three signs occurs.

Proof Positive real numbers in the old and new senses correspond
Strategy

Go back and forth between a section of the positive rationals and a section of all rationals by adding or removing the non-positive rationals from the lower class, and check that each direction gives a legitimate section.

Old to new

Let (L,R)(L, R) cut the positive rationals, LL with no greatest member. Put L′=L∪{x:x≤0}L' = L \cup \{x : x \le 0\}, R′=RR' = R. Every rational is now in exactly one class; L′L' lies below R′R' because non-positive numbers lie below the positive members of RR; L′L' has no greatest member since LL is non-empty and has none. And L′L' contains the positive members of LL, so the new section is positive.

New to old

Conversely, let (L′,R′)(L', R') be a positive real number. Then 0∈L′0 \in L', so R′R' consists of positive rationals only. Removing the non-positive rationals from L′L' leaves a non-empty set LL of positive rationals (the number is positive) with no greatest member, and (L,R′)(L, R') is a section of the positive rationals.

The two operations undo each other

Adding and then removing the non-positive rationals gives back LL; removing and then adding gives back L′L', since L′L' already contained every non-positive rational. So the correspondence is one-to-one.

Why not +∞+\infty and −∞-\infty?

We insisted that both classes be non-empty. Two “sections” are left out: the one with every rational in LL and the one with every rational in RR. Hardy remarks in a footnote that one could call these +∞+\infty and −∞-\infty, but that addition and multiplication then refuse to work sensibly (what would (+∞)+(−∞)(+\infty) + (-\infty) be?), and that a beginner has enough to do learning what “tends to infinity” means without meeting new numbers called infinity. In this book, ∞\infty is never a number.

Worked examples and exercises

Exercise The section at −1-1 ●○○ Hardy: §8

Show that L={x:x+1<0}L = \{x : x + 1 < 0\}, R={x:x+1≥0}R = \{x : x + 1 \ge 0\} is a real number, that it is negative, and that it corresponds to the rational −1-1.

Hint
Check the three requirements: both classes non-empty, LL below RR, LL with no greatest member.

Solution

−2∈L-2 \in L and 0∈R0 \in R, so both classes exist. If x∈Lx \in L and y∈Ry \in R then x<−1≤yx < -1 \le y. If x∈Lx \in L, then 12(x−1)\tfrac12(x - 1) is rational and lies strictly between xx and −1-1, so it is a larger member of LL: no greatest member. So the section is a real number. Its upper class contains the negative rational −1-1, so it is negative. And −1-1 is the least member of RR, so the section corresponds to the rational −1-1, with −1-1 in the upper class as the convention requires.

Exercise A negative irrational ●●○ Hardy: §8

Show that L={x:x3<−2}L = \{x : x^3 < -2\}, R={x:x3>−2}R = \{x : x^3 > -2\} is a negative irrational number.

Hint
No rational has cube −2-2 (if (p/q)3=−2(p/q)^3 = -2 then (−p/q)3=2(-p/q)^3 = 2), so every rational is in exactly one class.
Hint
To see that LL has no greatest member, compare with the section for 23\sqrt[3]{2}: x∈Lx \in L exactly when −x-x is in the upper class of 23\sqrt[3]2.

Solution

Since (p/q)3=−2(p/q)^3 = -2 would give (−p/q)3=2(-p/q)^3 = 2, which is impossible (Examples II, 1), every rational is in exactly one class. Cubing is increasing, so if x3<−2<y3x^3 < -2 < y^3 then x<yx < y: LL lies below RR. Both are non-empty (−2∈L-2 \in L, 0∈R0 \in R). The map x↦−xx \mapsto -x carries LL onto the upper class {y:y3>2}\{y : y^3 > 2\} of the section for 23\sqrt[3]2, reversing order. That class has no least member (the argument of the last lesson works for cubes with (x3−2)(x^3 - 2) in place of (x2−2)(x^2 - 2)), so LL has no greatest member, and similarly RR has no least member. So the section is an irrational real number, and since −1∈R-1 \in R is negative, it is negative. It is the number we write −23-\sqrt[3]{2}.

Exercise The lower class is enough ●●○ Hardy: §8

Hardy remarks that we could just as well have defined a real number to be its lower class alone. Show that the lower class determines the section, and describe which sets of rationals are lower classes.

Hint
Given LL, what must RR be?

Solution

Every rational is in exactly one class, so RR is the set of rationals not in LL: the lower class determines the upper class. A set LL of rationals is the lower class of a real number exactly when: (i) LL is non-empty and not all of the rationals; (ii) LL is closed downwards (if x∈Lx \in L and y<xy < x then y∈Ly \in L); (iii) LL has no greatest member. Indeed (ii) is what “every member of LL is less than every member of RR” says when RR is the complement of LL. This is the definition of a Dedekind cut found in many modern books.

Exercise Why the infinite “sections” are excluded ●●○ Hardy: §8, footnote

Suppose we allowed the section with LL = all rationals and called it +∞+\infty, and the one with RR = all rationals and called it −∞-\infty. Try to define (+∞)+(−∞)(+\infty) + (-\infty) by the rule for sums used later (lower class: all sums a+ba + b of lower members). What do you get?

Hint
The lower class of −∞-\infty is empty.

Solution

The lower class of the sum would consist of all a+ba + b with aa in the lower class of +∞+\infty and bb in the lower class of −∞-\infty. The second class is empty, so the set of sums is empty, while the upper class (all A+BA + B) would be empty too, since +∞+\infty has no upper members. So the rule produces no section at all. Patching such cases one by one is possible but awkward, which is Hardy’s point: it is simpler to keep ∞\infty out of the number system.

Check yourself

Check yourself Choose one

A real number’s lower class contains the rational −110-\tfrac{1}{10} but not 00. What can you conclude?

Why

00 is in the upper class, so the number is not positive. It is zero if every negative rational is in the lower class, but it could also be, say, −120-\tfrac{1}{20}, whose upper class contains −120-\tfrac{1}{20} and −130-\tfrac{1}{30}. One negative member of LL decides nothing.

Check yourself True or false?

The sections {x:x<12}\{x : x < \tfrac12\} and {x:x≤12}\{x : x \le \tfrac12\} (with the rest in the upper class) are two different real numbers.

Why

False. The second has a greatest member 12\tfrac12 in its lower class, so it is not a real number in Hardy’s sense at all; the convention that the lower class has no greatest member keeps exactly one section for each rational.

Check yourself True or false?

Under Hardy’s definitions, the symbol 12\tfrac12 can mean either a fraction or a section, and statements such as 12>13\tfrac12 > \tfrac13 are true in both readings.

Why

True. Hardy discusses this ambiguity and calls it harmless: the order and arithmetic of the sections corresponding to rationals agree with the order and arithmetic of the rationals themselves, so every inference goes through in either reading.

Hardy’s language

What is essential in mathematics is that its symbols should be capable of some interpretation; generally they are capable of many, and then, so far as mathematics is concerned, it does not matter which we adopt.

— G. H. Hardy, §8

This is why the choice “a real number is a pair of classes” is not as strange as it first sounds. What matters is that the sections behave exactly as numbers should; nobody is claiming that this is what 2\sqrt2 “really is”.

Summary

Summary
  • A real number is a section of all the rationals into non-empty classes LL below RR, with LL having no greatest member.
  • It is rational if RR has a least member (that rational), irrational otherwise.
  • Its sign is read off from 00: positive if 0∈L0 \in L; zero if LL is exactly the negatives; negative if some negative rational is in RR. Exactly one case occurs.
  • The positive reals of §7 are recovered by adding 00 and the negatives to the lower class; “sections” with an empty class are not numbers, so ±∞\pm\infty are not numbers.