Skip to content

§2 The Set ℚ of Rational Numbers

Subtraction forces us from N\mathbb{N} to the integers Z={0,1,−1,2,−2,… }\mathbb{Z} = \{0, 1, -1, 2, -2, \dots\}, and division forces us on to the rational numbers

Q={mn:m,n∈Z, n≠0}.\mathbb{Q} = \left\{ \tfrac{m}{n} : m, n \in \mathbb{Z},\ n \ne 0 \right\}.

Q\mathbb{Q} is a very satisfactory system for addition, subtraction, multiplication and division. It breaks down when we try to solve equations like x2=2x^2 = 2. A unit square has a diagonal dd with d2=12+12=2d^2 = 1^2 + 1^2 = 2, so there “ought” to be a number 2\sqrt2. We will see that no rational number works, even though rationals like 1.41421.4142 and 1.41431.4143 come very close: their squares are 1.999961641.99996164 and 2.000244492.00024449.

So Q\mathbb{Q} has gaps. The graph of y=x2−2y = x^2 - 2 crosses the xx-axis, yet it slips between all the rational points on that axis. Numbers like π\pi and ee also fall into these gaps.

Why this matters

This section is not really about algebraic numbers for their own sake. Its job is to show, beyond doubt, that Q\mathbb{Q} has holes, which is why analysis needs a bigger number system with no holes at all (§4). The tool for this is the Rational Zeros Theorem below. It turns an infinite search (“is there any fraction whose square is 22?”) into a short, finite list of candidates to check.

2.1 Definition.

A number is called an algebraic number if it satisfies a polynomial equation

cnxn+cn−1xn−1+⋯+c1x+c0=0c_n x^n + c_{n-1} x^{n-1} + \cdots + c_1 x + c_0 = 0

where the coefficients c0,c1,…,cnc_0, c_1, \dots, c_n are integers, cn≠0c_n \ne 0 and n≥1n \ge 1.

Used in §16

Every rational r=mnr = \frac mn is algebraic, since it solves nx−m=0nx - m = 0. Expressions built from rationals with roots  , 3,…\sqrt{\ }, \sqrt[3]{\ }, \dots and ordinary arithmetic are algebraic as well.

Example 1.

3\sqrt{3}, 417\tfrac{4}{17}, 173\sqrt[3]{17}, 2+53\sqrt{2 + \sqrt[3]{5}} and 4−237\sqrt{\tfrac{4 - 2\sqrt3}{7}} are algebraic. The first three solve x2−3=0x^2 - 3 = 0, 17x−4=017x - 4 = 0 and x3−17=0x^3 - 17 = 0.

If a=2+53a = \sqrt{2 + \sqrt[3]5}, then a2−2=53a^2 - 2 = \sqrt[3]5, so (a2−2)3=5(a^2 - 2)^3 = 5. Expanding, aa satisfies

x6−6x4+12x2−13=0.x^6 - 6x^4 + 12x^2 - 13 = 0.

If b=(4−23)/7b = \sqrt{(4 - 2\sqrt3)/7}, then 7b2=4−237b^2 = 4 - 2\sqrt3, so 23=4−7b22\sqrt3 = 4 - 7b^2 and 12=(4−7b2)212 = (4 - 7b^2)^2. So bb satisfies 49x4−56x2+4=049x^4 - 56x^2 + 4 = 0.

The next theorem tells us exactly which rationals could possibly solve such an equation. It lets us prove that algebraic numbers that look irrational usually are. If the statement seems complicated, read Corollary 2.3 and Examples 2–5 first.

2.2 Rational Zeros Theorem.

Suppose c0,c1,…,cnc_0, c_1, \dots, c_n are integers and rr is a rational number satisfying

cnxn+cn−1xn−1+⋯+c1x+c0=0(1)c_n x^n + c_{n-1} x^{n-1} + \cdots + c_1 x + c_0 = 0 \tag{1}

where n≥1n \ge 1, cn≠0c_n \ne 0 and c0≠0c_0 \ne 0. Write r=cdr = \frac cd where c,dc, d are integers with no common factors and d≠0d \ne 0. Then cc divides c0c_0 and dd divides cnc_n.

Used in Corollary 2.3, §4, Exercise 6.5

In other words, the only rational candidates for solutions of (1) have the form cd\frac cd where cc divides c0c_0 and dd divides cnc_n.

Proof

We are given cn(cd)n+cn−1(cd)n−1+⋯+c1cd+c0=0c_n \left(\frac cd\right)^n + c_{n-1} \left(\frac cd\right)^{n-1} + \cdots + c_1 \frac cd + c_0 = 0. Multiplying through by dnd^n,

cncn+cn−1cn−1d+cn−2cn−2d2+⋯+c1c dn−1+c0dn=0.(2)c_n c^n + c_{n-1} c^{n-1} d + c_{n-2} c^{n-2} d^2 + \cdots + c_1 c\, d^{n-1} + c_0 d^n = 0. \tag{2}

Solving for c0dnc_0 d^n,

c0dn=−c [cncn−1+cn−1cn−2d+⋯+c1dn−1],c_0 d^n = -c\,\bigl[c_n c^{n-1} + c_{n-1} c^{n-2} d + \cdots + c_1 d^{n-1}\bigr],

so cc divides c0dnc_0 d^n. Since cc and dnd^n have no common factors, cc divides c0c_0. (Here we use the basic fact that if cc divides abab and has no common factor with bb, then cc divides aa.)

Solving (2) for cncnc_n c^n instead,

cncn=−d [cn−1cn−1+cn−2cn−2d+⋯+c0dn−1],c_n c^n = -d\,\bigl[c_{n-1} c^{n-1} + c_{n-2} c^{n-2} d + \cdots + c_0 d^{n-1}\bigr],

so dd divides cncnc_n c^n. Since dd and cnc^n have no common factors, dd divides cnc_n.

∎

Intuitionwhy the proof works

After clearing denominators in (2), every term except c0dnc_0 d^n visibly contains a factor of cc. Since the terms add up to 00, which cc divides, the leftover term c0dnc_0 d^n must be divisible by cc too, and cc shares no factor with dd. The same balancing trick, run with dd instead of cc, handles cncnc_n c^n. This is the familiar “even/odd” proof that 2\sqrt2 is irrational, with divisibility by cc and dd in place of divisibility by 22.

2.3 Corollary.

Consider the polynomial equation xn+cn−1xn−1+⋯+c1x+c0=0x^n + c_{n-1}x^{n-1} + \cdots + c_1 x + c_0 = 0, where c0,…,cn−1c_0, \dots, c_{n-1} are integers and c0≠0c_0 \ne 0. Any rational solution of this equation must be an integer that divides c0c_0.

Used in Exercise 2.1, Exercise 2.2, Exercise 2.8, Exercise 4.12

Proof

In Theorem 2.2 the denominator dd must divide the leading coefficient, which is 11. So rr is an integer, and it divides c0c_0.

∎

Polynomials whose leading coefficient is 11 are called monic.

Enter integer coefficients from the highest power down. The widget lists every candidate ±c/d\pm c/d allowed by Theorem 2.2 and evaluates the polynomial at each one exactly, using integer arithmetic.

6x³ − 11x² + 6x − 1 = 0

Divisors of c₀ = -1: 1. Divisors of cₙ = 6: 1, 2, 3, 6. So the only possible rational roots are the 8 numbers ±c/d below.

-1P = -24
-1/2P = -15/2
-1/3P = -40/9
-1/6P = -7/3
1/6P = -5/18
1/3root ✓
1/2root ✓
1root ✓

Rational roots: 1/3, 1/2, 1.

Example 2.

2\sqrt2 is not a rational number.

Proof

By Corollary 2.3 the only possible rational solutions of x2−2=0x^2 - 2 = 0 are ±1,±2\pm1, \pm2. None of these four numbers works. Since 2\sqrt2 is a solution, it is not rational.

∎
Example 3.

17\sqrt{17} is not a rational number.

Proof

The only possible rational solutions of x2−17=0x^2 - 17 = 0 are ±1,±17\pm1, \pm17, and none of them is a solution.

∎
Example 4.

63\sqrt[3]{6} is not a rational number.

Proof

The only possible rational solutions of x3−6=0x^3 - 6 = 0 are ±1,±2,±3,±6\pm1, \pm2, \pm3, \pm6, and none of these eight numbers works.

∎
Example 5.

a=2+53a = \sqrt{2 + \sqrt[3]{5}} is not a rational number.

Proof

By Example 1, aa solves x6−6x4+12x2−13=0x^6 - 6x^4 + 12x^2 - 13 = 0, whose only possible rational solutions are ±1,±13\pm1, \pm13. At x=±1x = \pm1 the left side is −6-6. At x=±13x = \pm13 we can avoid the arithmetic: aa is obviously between 11 and 1313. Alternatively, 136−6⋅134+12⋅132−13=13 (135−6⋅133+12⋅13−1)≠013^6 - 6\cdot13^4 + 12\cdot13^2 - 13 = 13\,(13^5 - 6\cdot13^3 + 12\cdot13 - 1) \ne 0, since the term in parentheses is one less than a multiple of 1313.

∎
Example 6.

b=(4−23)/7b = \sqrt{(4 - 2\sqrt3)/7} is not a rational number.

Proof

By Theorem 2.2 the possible rational roots of 49x4−56x2+4=049x^4 - 56x^2 + 4 = 0 are ±1,±17,±149,±2,±27,±249,±4,±47,±449\pm1, \pm\frac17, \pm\frac1{49}, \pm2, \pm\frac27, \pm\frac2{49}, \pm4, \pm\frac47, \pm\frac4{49}. Checking all 18 is discouraging, so here is a cleverer route. Example 1 showed 12=(4−7b2)212 = (4 - 7b^2)^2. If bb were rational, so would be 4−7b24 - 7b^2 (Exercise 2.6), and then x2=12x^2 = 12 would have a rational solution. But the only candidates ±1,±2,±3,±4,±6,±12\pm1, \pm2, \pm3, \pm4, \pm6, \pm12 all fail. So bb is irrational.

∎

In practice, estimating the number in question eliminates most candidates at once. The numbers in Examples 2–5 are clearly not integers. In Example 6, b≈0.2767b \approx 0.2767, and the nearest candidate is 27≈0.2857\frac27 \approx 0.2857. Be careful, though: not every irrational-looking expression is irrational. See Exercise 2.7.

Common pitfall

  • The theorem produces candidates. It never says that any of them is actually a root; you still have to check each one.
  • It needs c0≠0c_0 \ne 0. If the constant term is 00, factor out a power of xx first: x3−2x=x(x2−2)x^3 - 2x = x(x^2 - 2), and apply the theorem to x2−2x^2 - 2.
  • The fraction must be in lowest terms. The number 24\frac24 is a root of 2x−1=02x - 1 = 0, yet 22 does not divide c0=−1c_0 = -1. Writing it as 12\frac12 fixes this.
  • To prove a number is irrational this way, you first need an integer polynomial that it satisfies. Finding that polynomial, as in Example 1, is often the real work.
2.4 Remark.

The Rational Zeros Theorem finds the rational zeros of integer polynomials. To approximate other zeros, of these or other functions, one uses Newton’s method and the secant method, discussed in §31.

Reading alongside Abbott

Abbott opens with the classical proof that 2\sqrt2 is irrational in Section 1.1: assume 2=pq\sqrt2 = \frac pq in lowest terms and show that pp and qq would both be even. Theorem 2.2 is that argument made to work for every integer polynomial at once. In Section 1.3, Abbott uses a set of rationals with no rational least upper bound to motivate completeness. Here that idea appears in §4, right after the Completeness Axiom.

Take the quiz (3 questions)

By Corollary 2.3, which numbers are the only possible rational solutions of x2−6=0x^2 - 6 = 0?

Suppose r=cdr = \frac cd in lowest terms solves cnxn+⋯+c1x+c0=0c_n x^n + \cdots + c_1 x + c_0 = 0 with integer coefficients and c0≠0c_0 \ne 0. What does the Rational Zeros Theorem say?

Is every algebraic number rational?

Exercise 2.1

Show 3\sqrt3, 5\sqrt5, 7\sqrt7, 24\sqrt{24} and 31\sqrt{31} are not rational numbers.

Hint

Apply Corollary 2.3 to x2−k=0x^2 - k = 0. In each case the candidates are the divisors of kk, and none of them squares to kk.

Solution

Suppose k\sqrt k were rational. It would be a rational solution of x2−k=0x^2 - k = 0, so by Corollary 2.3 it would be an integer dividing kk. We check that no divisor dd of kk (positive or negative) has d2=kd^2 = k.

  • k=3k = 3: divisors ±1,±3\pm1, \pm3 have squares 1,91, 9.
  • k=5k = 5: divisors ±1,±5\pm1, \pm5 have squares 1,251, 25.
  • k=7k = 7: divisors ±1,±7\pm1, \pm7 have squares 1,491, 49.
  • k=24k = 24: divisors ±1,±2,±3,±4,±6,±8,±12,±24\pm1, \pm2, \pm3, \pm4, \pm6, \pm8, \pm12, \pm24 have squares 1,4,9,16,36,64,144,5761, 4, 9, 16, 36, 64, 144, 576.
  • k=31k = 31: divisors ±1,±31\pm1, \pm31 have squares 1,9611, 961.

None equals kk, so none of these square roots is rational.

Tags: StandardContradiction

Exercise 2.2

Show 23\sqrt[3]{2}, 57\sqrt[7]{5} and 134\sqrt[4]{13} are not rational numbers.

Hint

The candidates for x3−2x^3 - 2 are ±1,±2\pm1, \pm2; for x7−5x^7 - 5 they are ±1,±5\pm1, \pm5; for x4−13x^4 - 13 they are ±1,±13\pm1, \pm13.

Solution

Each number is a root of xn−k=0x^n - k = 0 for the indicated n,kn, k, so a rational one would be an integer dividing kk (Corollary 2.3).

  • 23\sqrt[3]2: x3−2x^3 - 2; candidates ±1,±2\pm1, \pm2 have cubes ±1,±8≠2\pm1, \pm8 \neq 2.
  • 57\sqrt[7]5: x7−5x^7 - 5; candidates ±1,±5\pm1, \pm5 have seventh powers ±1,±78125≠5\pm1, \pm78125 \neq 5.
  • 134\sqrt[4]{13}: x4−13x^4 - 13; candidates ±1,±13\pm1, \pm13 have fourth powers 1,28561≠131, 28561 \neq 13.

So none of them is rational.

Tags: StandardContradiction

Exercise 2.3

Show 2+2\sqrt{2 + \sqrt2} is not a rational number.

Solution

If a=2+2a = \sqrt{2 + \sqrt2} then (a2−2)2=2(a^2 - 2)^2 = 2, so aa solves x4−4x2+2=0x^4 - 4x^2 + 2 = 0. The candidates are ±1,±2\pm1, \pm2, giving values −1-1 and 22. None is zero.

Tags: StandardContradiction

Exercise 2.4

Show 5−33\sqrt[3]{5 - \sqrt3} is not a rational number.

Solution

If a=5−33a = \sqrt[3]{5 - \sqrt3} then (5−a3)2=3(5 - a^3)^2 = 3, so aa solves x6−10x3+22=0x^6 - 10x^3 + 22 = 0. The candidates are ±1,±2,±11,±22\pm1, \pm2, \pm11, \pm22. Since 1<a<21 < a < 2, none of them is aa.

Tags: StandardContradiction

Exercise 2.5

Show [3+2]2/3\left[3 + \sqrt2\right]^{2/3} is not a rational number.

Solution

If a=(3+2)2/3a = (3 + \sqrt2)^{2/3} then a3=(3+2)2=11+62a^3 = (3 + \sqrt2)^2 = 11 + 6\sqrt2, so (a3−11)2=72(a^3 - 11)^2 = 72 and aa solves x6−22x3+49=0x^6 - 22x^3 + 49 = 0. The candidates are ±1,±7,±49\pm1, \pm7, \pm49, but a3≈19.5a^3 \approx 19.5 gives a≈2.69a \approx 2.69.

Tags: StandardContradiction

Exercise 2.6

In connection with Example 6, discuss why 4−7b24 - 7b^2 is rational if bb is rational.

Solution

Q\mathbb{Q} is closed under multiplication and subtraction: if b=mnb = \frac mn then 4−7b2=4n2−7m2n24 - 7b^2 = \frac{4n^2 - 7m^2}{n^2}.

Tags: Warm-upExplain whyComputation

Exercise 2.7

Show the following irrational-looking expressions are actually rational numbers.

  1. 4+23−3\sqrt{4 + 2\sqrt3} - \sqrt3
  2. 6+42−2\sqrt{6 + 4\sqrt2} - \sqrt2
Solution

(a) 4+23=(1+3)24 + 2\sqrt3 = (1 + \sqrt3)^2, so the expression is 11. (b) 6+42=(2+2)26 + 4\sqrt2 = (2 + \sqrt2)^2, so the expression is 22.

Tags: Warm-upComputation

Exercise 2.8

Find all rational solutions of the equation x8−4x5+13x3−7x+1=0x^8 - 4x^5 + 13x^3 - 7x + 1 = 0.

Solution

By Corollary 2.3 the only candidates are ±1\pm1. At x=1x = 1: 1−4+13−7+1=41 - 4 + 13 - 7 + 1 = 4. At x=−1x = -1: 1+4−13+7+1=01 + 4 - 13 + 7 + 1 = 0. So x=−1x = -1 is the only rational solution. Try it in the widget above.

Tags: StandardComputationContradiction

Keyboard shortcuts

Single-key shortcuts
KeyAction
nNext section
pPrevious section
jNext exercise
kPrevious exercise
hShow or hide all hints
sShow or hide all solutions
/Search
?Show this list
Ctrl K or ⌘ KSearch (built in)
EscClose a dialog

Shortcuts never fire while you type in a text field or hold Ctrl, Alt or ⌘. Screen reader users in browse mode can press the keys after switching to focus mode, or turn them off here.