Chapter I · Real variables

The continuum and the real variable

§§15–16 30 min

How few real numbers are surds, how a section defines numbers that no formula can express, and what Hardy means by a variable.

Why this matters

All the irrationals met so far, 2\sqrt2, 1+31 + \sqrt3, 2+2\sqrt{2 + \sqrt2}, have formulas. It is tempting to think every real number does. §15 shows how wrong that is: some roots of polynomial equations have no formula in surds, and numbers such as π\pi are not roots of any polynomial with integer coefficients at all. Sections handle all of them equally well, because a section needs no formula, only a test that sorts rationals into left and right. §16 then introduces the language used in the rest of the book: the variable xx and its field of values.

The idea

The collection of all real numbers is the arithmetical continuum. Hardy assumes that the points of his line correspond exactly to the real numbers, the linear continuum, so that we can keep drawing pictures.

Consider the equation z3=3z+8z^3 = 3z + 8. Cardan’s formula for cubics says its root is

z=4+153+4−153.z = \sqrt[3]{4 + \sqrt{15}} + \sqrt[3]{4 - \sqrt{15}}.

But there is a more direct way to define that root, one that works even when no formula exists: sort the rationals by the test x3<3x+8x^3 < 3x + 8 or x3>3x+8x^3 > 3x + 8.

Explore Numbers defined by a test, not a formula

Try this. With z3=3z+8z^3 = 3z + 8, add decimal places and compare the bracket with Cardan's value in the read-out. Test the rationals 22 and 52\tfrac52. Then switch to x5=x+16x^5 = x + 16 and do the same.

Equation
L={x:x3<3x+8}L=\{x : x^3 < 3x + 8\}, R = all other rationals
LR2.47502.48002.48502.49002.49502.50002.50502.51002.51502.492.50
  • L (lower class)
  • R (upper class)
  • gap not yet sorted
  • your fraction
With 2 decimal places: 2.49 is in L and 2.50 is in R; the gap between them has width 0.01. Your fraction 1712\tfrac{17}{12} ≈ 1.4166667 is in L.
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 section needs no formula: a rational xx goes into L exactly when x3<3x+8x^3 < 3x + 8, a test done in exact integer arithmetic. Cardan's formula gives the same number, 4+153+4−153≈2.492033301\sqrt[3]{4+\sqrt{15}} + \sqrt[3]{4-\sqrt{15}} \approx 2.492033301; compare it with the decimals above.
What to notice

For the cubic the bracket closes on 2.49203…2.49203\ldots, the value of Cardan's surd expression: the section and the formula agree. For the quintic the section works in exactly the same way, but there is no formula to compare with: no expression in rationals and roots of any order equals this number. The section is the only description we need.

So the real numbers come in increasingly exclusive kinds: the rationals; the quadratic surds and other numbers built from roots; the algebraic numbers, roots of polynomial equations with integer coefficients (most of which cannot be written with roots); and the transcendental numbers, which are not algebraic at all, such as π\pi. The special kinds are, in Hardy’s words, a minute fraction of the continuum.

Precise statements

Definition The continuum Hardy: §15

The set of all real numbers, rational and irrational, is the arithmetical continuum. Taking the points of the line Λ\Lambda to correspond to all real numbers and no others, we obtain the linear continuum.

Theorem The real root of z3=3z+8z^3 = 3z + 8 Hardy: §15 (i)

The equation z3=3z+8z^3 = 3z + 8 has exactly one real root. It is positive and irrational, and it is the positive real number defined by the section x3<3x+8x^3 < 3x + 8, x3>3x+8x^3 > 3x + 8 of the positive rationals.

Definition Algebraic and transcendental numbers Hardy: Misc. Examples I, 36

A number is algebraic if it is a root of an equation a0xn+a1xn−1+⋯+an=0a_0x^n + a_1x^{n-1} + \dots + a_n = 0 with integer coefficients, not all zero. An irrational number that is not algebraic is transcendental.

Fact Some numbers are hard to describe Hardy: §15 (ii), (iii)

The positive root of x5=x+16x^5 = x + 16 is algebraic, but cannot be expressed by rationals and roots of any order. The number π\pi is transcendental (Lindemann, 1882). Neither proof belongs to this course.

Definition Variable and field of variation Hardy: §16

A variable is a symbol xx standing for an unspecified member of a given set, its field of variation; the members of the set are the values of the variable. When the field is the set of all real numbers, xx is the continuous real variable; when it is the set of positive integers, nn is a positive integral variable.

Proofs

Hardy’s proof of the theorem shows how to establish a number’s existence directly, by a section, without any formula.

Proof z3=3z+8z^3 = 3z + 8 has exactly one real root, and the section defines it
Strategy

First rule out negative roots and a second positive root by inequalities. Then show that the root is not rational. Finally check that the test x3<3x+8x^3 < 3x + 8 really does split the positive rationals into a lower and an upper class.

No root is negative or zero

z=0z = 0 gives 0=80 = 8. If z=−ζz = -\zeta with ζ>0\zeta > 0, the equation becomes ζ3−3ζ+8=0\zeta^3 - 3\zeta + 8 = 0, that is, 3−ζ2=8/ζ3 - \zeta^2 = 8/\zeta. The right side is positive, so ζ2<3\zeta^2 < 3 and ζ<2\zeta < 2; but then 8/ζ>4>3>3−ζ28/\zeta > 4 > 3 > 3 - \zeta^2. Contradiction.

Any root exceeds 22

A root zz is positive, so z3=3z+8>8z^3 = 3z + 8 > 8 and z>2z > 2.

At most one root

If z1≠z2z_1 \neq z_2 were both roots, subtracting gives z13−z23=3(z1−z2)z_1^3 - z_2^3 = 3(z_1 - z_2); dividing by z1−z2z_1 - z_2, z12+z1z2+z22=3z_1^2 + z_1z_2 + z_2^2 = 3. But each of z1z_1, z2z_2 exceeds 22, so the left side exceeds 1212.

The root is not rational

By Gauss’s theorem (Examples II, 3) a rational root would be an integer dividing 88, and positive: 1,2,4,81, 2, 4, 8. None works (1−3−81 - 3 - 8, 8−6−88 - 6 - 8, 64−12−864 - 12 - 8, 512−24−8512 - 24 - 8 are all non-zero).

The test gives a section

If xx is in the upper class (x3>3x+8x^3 > 3x + 8, so x>2x > 2) and y>xy > x, then

(y3−3y)−(x3−3x)=(y−x)(y2+xy+x2−3)>0,(y^3 - 3y) - (x^3 - 3x) = (y - x)(y^2 + xy + x^2 - 3) > 0,

so yy is in the upper class too. Similarly, everything below a member of the lower class is in the lower class. Both classes are non-empty (11 and 33), and no rational is unclassified (the root is not rational). So the test defines a positive real number zz, and an argument like that of §12 shows z3=3z+8z^3 = 3z + 8.

Worked examples and exercises

Exercise Cardan's formula checks out ●●○ Hardy: §15 (i)

Let u=4+153u = \sqrt[3]{4 + \sqrt{15}} and v=4−153v = \sqrt[3]{4 - \sqrt{15}}. Show that z=u+vz = u + v satisfies z3=3z+8z^3 = 3z + 8.

Hint
Compute uvuv and u3+v3u^3 + v^3 first.
Hint
(u+v)3=u3+v3+3uv(u+v)(u + v)^3 = u^3 + v^3 + 3uv(u + v).

Solution

u3+v3=(4+15)+(4−15)=8u^3 + v^3 = (4 + \sqrt{15}) + (4 - \sqrt{15}) = 8, and (uv)3=(4+15)(4−15)=16−15=1(uv)^3 = (4 + \sqrt{15})(4 - \sqrt{15}) = 16 - 15 = 1, so uv=1uv = 1 (the only real cube root of 11). Hence

z3=(u+v)3=u3+v3+3uv(u+v)=8+3z.z^3 = (u + v)^3 = u^3 + v^3 + 3uv(u + v) = 8 + 3z.

Numerically u≈1.98936u \approx 1.98936 and v≈0.50267v \approx 0.50267, so z≈2.49203z \approx 2.49203, which lies in the bracket the section gives.

Exercise The quintic has one real root ●●○ Hardy: §15 (ii)

Show that x5=x+16x^5 = x + 16 has exactly one real root, and that it lies between 11 and 22.

Hint
Study g(x)=x5−xg(x) = x^5 - x on the intervals x≤−1x \le -1, −1<x<0-1 < x < 0, 0≤x≤10 \le x \le 1 and x>1x > 1.

Solution

We need g(x)=x5−x=16g(x) = x^5 - x = 16. For x≤−1x \le -1: g(x)=x(x4−1)≤0g(x) = x(x^4 - 1) \le 0. For −1<x<0-1 < x < 0: g(x)=x(x4−1)g(x) = x(x^4 - 1) is a product of two negatives, positive but less than ∣x∣<1|x| < 1. For 0≤x≤10 \le x \le 1: g(x)≤0g(x) \le 0. So no root has x≤1x \le 1. For 1<x<y1 < x < y: g(y)−g(x)=(y−x)(y4+y3x+y2x2+yx3+x4−1)>0g(y) - g(x) = (y - x)(y^4 + y^3x + y^2x^2 + yx^3 + x^4 - 1) > 0, since the bracket exceeds 5−15 - 1; so gg is increasing on x>1x > 1 and takes the value 1616 at most once. Finally g(1)=0<16<30=g(2)g(1) = 0 < 16 < 30 = g(2), and the section x5<x+16x^5 < x + 16, x5>x+16x^5 > x + 16 (whose classes are closed downwards and upwards by the same monotonicity) defines a number between 11 and 22 that is the root.

Exercise 2+3\sqrt2 + \sqrt3 is algebraic ●●○ Hardy: Misc. Examples I, 36–37

Find an equation with integer coefficients satisfied by x=2+3x = \sqrt2 + \sqrt3.

Hint
Square once, isolate the remaining square root, and square again.

Solution

x2=5+26x^2 = 5 + 2\sqrt6, so x2−5=26x^2 - 5 = 2\sqrt6 and (x2−5)2=24(x^2 - 5)^2 = 24, that is, x4−10x2+1=0x^4 - 10x^2 + 1 = 0. So 2+3\sqrt2 + \sqrt3 is algebraic. Hardy’s Example 37 says much more: sums, differences, products and quotients of algebraic numbers are always algebraic.

Exercise The field of variation matters ●○○ Hardy: §16

Decide whether each statement is true when the variable ranges over (a) the positive integers, (b) the rationals, (c) the real numbers: (1) “x2≠2x^2 \neq 2”; (2) “between xx and x+1x + 1 there is another value of the variable”; (3) “xx has a successor: a least value greater than xx”.

Hint
Statement (1) is about irrational numbers; (2) and (3) are about gaps between values.

Solution

(1) True for (a) and (b), since no rational squares to 22; false for (c), where x=2x = \sqrt2 is a value. (2) False for (a): nothing lies strictly between nn and n+1n + 1. True for (b) and (c): x+12x + \tfrac12 lies between. (3) True for (a): the successor of nn is n+1n + 1. False for (b) and (c): if y>xy > x then 12(x+y)\tfrac12(x + y) is a smaller value still greater than xx. The same words, “for every value of xx”, say different things for different fields.

Check yourself

Check yourself Choose one

Which description fits the positive root of x5=x+16x^5 = x + 16?

Why

It is a root of a polynomial with integer coefficients, so it is algebraic, hence not transcendental. It is not rational (the only candidates ±1,±2,±4,±8,±16\pm1, \pm2, \pm4, \pm8, \pm16 fail). Hardy states, without proof, that no expression in roots gives it; this is the theorem of Abel and Galois on quintics, and it holds for this particular equation.

Check yourself True or false?

Every real number can be written by a finite formula using integers, ++, −-, ×\times, ÷\div and roots.

Why

False. Such formulas give only certain algebraic numbers. Most algebraic numbers of degree 55 or more, and all transcendental numbers such as π\pi, have no such formula. Sections define all real numbers alike.

Check yourself Choose one

For which variable is the statement “there is a value of the variable whose square is 22” true?

Why

Only the field of all real numbers contains 2\sqrt2. This is the whole story of the chapter in one line: the move from the rational variable to the continuous variable is the move from a field with gaps to one without.

Hardy’s language

they comprise only a minute fraction of the infinite variety of numbers which constitute the continuum.

— G. H. Hardy, §15

Hardy means the rationals and the surds. The precise modern form of the remark is Cantor’s: the algebraic numbers can be listed in a single sequence (like the rationals in the first lesson), while the real numbers cannot.

Summary

Summary
  • The arithmetical continuum is the set of all real numbers; Hardy pictures it as the points of a line.
  • A section defines a number by a test on rationals, with no formula needed: z3=3z+8z^3 = 3z + 8 and x5=x+16x^5 = x + 16 each have exactly one real root, so defined.
  • Rationals, surds, algebraic numbers and transcendental numbers are successively larger kinds; π\pi is transcendental, and most algebraic numbers have no surd formula.
  • A variable stands for an unspecified member of its field of variation; whether a statement about “any xx” is true depends on that field.