Gaps in a crowded line

The ratio­nals are dense, and yet, strangely, they still leave gaps. The diag­o­nal of a square of side one is a length that no frac­tion can name. We will prove this, and then draw such a length on the num­ber line. After a short visit to π\pi, we turn to dec­i­mals. Which ratio­nals give ter­mi­nat­ing dec­i­mals? Why do the rest repeat? How do you turn a repeat­ing dec­i­mal back into a frac­tion? And how does the dec­i­mal of an irra­tional num­ber look dif­fer­ent from both?

A length with no frac­tion

A square of side 11 has a diag­o­nal dd with d2=1+1=2d^2 = 1 + 1 = 2, by the Baud­hāyana–Pythago­ras the­o­rem. So d=2d = \sqrt{2}. Baud­hāyana met lengths like this around 800 BCE while lay­ing out altars, and the Greeks came across them a few cen­turies later.

A square with sides of length 1 and its diagonal drawn in red; the diagonal d, the hypotenuse of a right triangle with legs 1 and 1, is marked d = root 2.
The diag­o­nal of a unit square has length root 2.

Num­bers on the line that are not ratios of inte­gers are called irra­tional. 2\sqrt{2} is one of them. The proof uses con­tra­dic­tion, a lovely method: you assume the oppo­site, and then watch it fall apart.

Claim. 2\sqrt{2} is not ratio­nal.

Sup­pose it is. Then 2=pq\displaystyle \sqrt{2} = \dfrac{p}{q} in low­est terms, which means pp and qq have no com­mon fac­tor.

Squar­ing both sides, 2=p2q2\displaystyle 2 = \dfrac{p^2}{q^2}, so p2=2q2p^2 = 2q^2.

That makes p2p^2 even. Since an odd num­ber squared is always odd, pp itself must be even, say p=2kp = 2k.

Put this back in: 4k2=2q24k^2 = 2q^2, so q2=2k2q^2 = 2k^2. Now q2q^2 is even, so qq is even too.

But look what has hap­pened. pp and qq are both even, so both are divis­i­ble by 22. That goes against "no com­mon fac­tor". Our assump­tion must have been false, and so 2\sqrt{2} is irra­tional.

The same argu­ment with 33 in place of 22 shows that 3\sqrt{3} is irra­tional. The same is true of 5\sqrt{5}, 7\sqrt{7}, and indeed n\sqrt{n} for any nn that is not a per­fect square.

Draw­ing 2\sqrt{2} on the num­ber line

You can actu­ally draw this num­ber. Mark OO at 00 and AA at 11. At AA, draw a per­pen­dic­u­lar and mark BB on it so that AB=1AB = 1. Then OB=1+1=2OB = \sqrt{1 + 1} = \sqrt{2}. Put the com­pass point at OO, open it to OBOB, and swing an arc down to the line. It lands exactly at 2\sqrt{2}.

Number line from 0 to 3 with O at 0 and A at 1; AB of length 1 stands perpendicular at A, OB is root 2, and a dashed arc centred at O carries B down to the point root 2.
Swing­ing OB down to the line marks root 2.

Now repeat from that point with another unit per­pen­dic­u­lar, and you get 3\sqrt{3}. Keep going for 4=2\sqrt{4} = 2, then 5\sqrt{5}, and so on. The result is a beau­ti­ful spi­ral of right tri­an­gles that gives you every n\sqrt{n}.

A spiral of right triangles, each with an outer side of 1 built on the previous hypotenuse, giving hypotenuses root 2, root 3, root 4 = 2, root 5 and root 6.
Each new unit per­pen­dic­u­lar gives the next square root.

π\pi

The ratio of a cir­cle's cir­cum­fer­ence to its diam­e­ter is π\pi. Āryab­haṭa gave 6283220000=3.1416\displaystyle \dfrac{62832}{20000} = 3.1416, and he was care­ful to call it an approx­i­ma­tion, not the exact value. He was right to be care­ful. π\pi is irra­tional (Lam­bert proved this in 1761), so no frac­tion is exact. In the four­teenth cen­tury, Mād­hava of Saṅga­m­a­grāma found an exact expres­sion that uses infi­nitely many terms.

π=4(1−13+15−17+⋯ )\displaystyle \pi = 4\left(1 - \frac{1}{3} + \frac{1}{5} - \frac{1}{7} + \cdots\right)

The more terms you add, the closer you get, and the value you are clos­ing in on is π\pi.

Real num­bers

Put the ratio­nals and irra­tionals together and they fill the whole line, with no gaps left. This com­plete set is called the real num­bers, R\mathbb{R}.

Dec­i­mals of ratio­nal num­bers

When you divide pp by qq, one of just two things can hap­pen.

It ter­mi­nates. 78=0.875\displaystyle \dfrac{7}{8} = 0.875. Here the remain­der reaches 00 and the divi­sion stops.

It repeats. 411=0.3636…=0.36‾\displaystyle \dfrac{4}{11} = 0.3636\ldots = 0.\overline{36}. Here the remain­der never reaches 00. But there are only q−1q - 1 pos­si­ble nonzero remain­ders, so sooner or later one of them must come back, and from then on the dig­its go round in a cycle.

Here is the nice part: you can pre­dict which will hap­pen with­out divid­ing at all. Write pq\displaystyle \dfrac{p}{q} in low­est terms and fac­torise qq. If qq has no prime fac­tors other than 22 and 55, the dec­i­mal ter­mi­nates, because the denom­i­na­tor can be scaled up to a power of 1010. For exam­ple, 940=923×5=9×251000=0.225\displaystyle \dfrac{9}{40} = \dfrac{9}{2^3 \times 5} = \dfrac{9 \times 25}{1000} = 0.225. Any other prime in qq makes it repeat. Take 512\displaystyle \dfrac{5}{12}. Here 12=22×312 = 2^2 \times 3, and that fac­tor of three makes it repeat: 0.416‾0.41\overline{6}.

From repeat­ing dec­i­mal back to frac­tion

Pure repeat­ing dec­i­mals, where the block starts right after the point, are the eas­ier kind. Call the num­ber xx, mul­ti­ply by 10n10^n where nn is the length of the block, and sub­tract.

Exam­ple 1. x=0.7‾x = 0.\overline{7}. 10x=7.7‾10x = 7.\overline{7}. Sub­tract: 9x=79x = 7, x=79\displaystyle x = \dfrac{7}{9}.

Exam­ple 2. x=0.27‾x = 0.\overline{27}. 100x=27.27‾100x = 27.\overline{27}. 99x=2799x = 27, x=2799=311\displaystyle x = \dfrac{27}{99} = \dfrac{3}{11}.

Mixed dec­i­mals have some dig­its before the block starts. First shift the non-repeat­ing dig­its past the point, then shift one full block more, and sub­tract.

Exam­ple 3. x=0.83‾x = 0.8\overline{3}. There is one non-repeat­ing digit, so 10x=8.3‾10x = 8.\overline{3}. The block has one digit, so 100x=83.3‾100x = 83.\overline{3}. Sub­tract­ing, 90x=7590x = 75, x=7590=56\displaystyle x = \dfrac{75}{90} = \dfrac{5}{6}.

Exam­ple 4. x=1.245‾x = 1.2\overline{45}. 10x=12.45‾10x = 12.\overline{45}; 1000x=1245.45‾1000x = 1245.\overline{45}. Sub­tract: 990x=1233990x = 1233, x=1233990=137110\displaystyle x = \dfrac{1233}{990} = \dfrac{137}{110}.

Exam­ple 5. Here is one that sur­prises every­body: why is 0.9‾=10.\overline{9} = 1? Let x=0.9‾x = 0.\overline{9}. Then 10x=9.9‾10x = 9.\overline{9}, so 9x=99x = 9 and x=1x = 1. Not approx­i­mately, but exactly. In fact every ter­mi­nat­ing dec­i­mal has a twin end­ing in repeat­ing nines, such as 3.25=3.249‾3.25 = 3.24\overline{9}.

A cyclic curios­ity

17=0.142857‾\displaystyle \dfrac{1}{7} = 0.\overline{142857}. Try mul­ti­ply­ing the block by 2,3,4,5,62, 3, 4, 5, 6. You get 285714285714, 428571428571, 571428571428, 714285714285 and 857142857142. They are the same six dig­its every time, just rotated. Repeat­ing blocks like this are called cyclic num­bers.

Dec­i­mals of irra­tional num­bers

These nei­ther ter­mi­nate nor repeat, as in 2=1.41421356…\sqrt{2} = 1.41421356\ldots and π=3.14159265…\pi = 3.14159265\ldots. Be care­ful, though: a dec­i­mal can fol­low a pat­tern and still be irra­tional, as long as the pat­tern never set­tles into a fixed block. In 0.101001000100001…0.101001000100001\ldots the runs of zeros keep get­ting longer, so no block ever repeats.

Prac­tice

  1. With­out divid­ing, say which ter­mi­nate: 1140\displaystyle \dfrac{11}{40}, 730\displaystyle \dfrac{7}{30}, 19125\displaystyle \dfrac{19}{125}, 314\displaystyle \dfrac{3}{14}. Then find the dec­i­mals.
  2. Write as frac­tions in low­est terms: (a) 0.4‾0.\overline{4}; (b) 0.81‾0.\overline{81}; (c) 0.52‾0.5\overline{2}; (d) 2.136‾2.1\overline{36}.
  3. Clas­sify as ratio­nal or irra­tional: (a) 49\sqrt{49}; (b) 20\sqrt{20}; (c) 0.123‾0.\overline{123}; (d) 0.12112111211112…0.12112111211112\ldots; (e) 227\displaystyle \dfrac{22}{7}; (f) π\pi.
  4. Prove that 5\sqrt{5} is irra­tional, fol­low­ing the steps used for 2\sqrt{2}. (Use: if p2p^2 is a mul­ti­ple of 55, so is pp.)
  5. Describe how to mark 10\sqrt{10} on the num­ber line using a right tri­an­gle with legs 33 and 11.
  6. Find the dec­i­mal expan­sions of 113\displaystyle \dfrac{1}{13} and 213\displaystyle \dfrac{2}{13}. What do you notice about the repeat­ing blocks?
  7. Is 227\displaystyle \dfrac{22}{7} equal to π\pi? Explain in one sen­tence.

Answers

Show answers
  1. 40=23⋅540 = 2^3 \cdot 5: ter­mi­nates, 0.2750.275. 3030 has a 33: repeats, 0.23‾0.2\overline{3}. 125=53125 = 5^3: ter­mi­nates, 0.1520.152. 1414 has a 77: repeats, 0.2142857‾0.2\overline{142857}.
  2. (a) 49\displaystyle \dfrac{4}{9}. (b) 8199=911\displaystyle \dfrac{81}{99} = \dfrac{9}{11}. (c) 10x=5.2‾10x = 5.\overline{2}, 100x=52.2‾100x = 52.\overline{2}, 90x=4790x = 47: 4790\displaystyle \dfrac{47}{90}. (d) 10x=21.36‾10x = 21.\overline{36}, 1000x=2136.36‾1000x = 2136.\overline{36}, 990x=2115990x = 2115: 2115990=4722\displaystyle \dfrac{2115}{990} = \dfrac{47}{22}.
  3. (a) Ratio­nal, 77. (b) Irra­tional (2020 is not a per­fect square). (c) Ratio­nal, 123999=41333\displaystyle \dfrac{123}{999} = \dfrac{41}{333}. (d) Irra­tional: the runs of 11s grow, so no block repeats. (e) Ratio­nal. (f) Irra­tional.
  4. Sup­pose 5=pq\displaystyle \sqrt{5} = \dfrac{p}{q} in low­est terms. Then p2=5q2p^2 = 5q^2, so p2p^2 and hence pp is a mul­ti­ple of 55: p=5kp = 5k. Then 25k2=5q225k^2 = 5q^2, q2=5k2q^2 = 5k^2, so qq is a mul­ti­ple of 55. Both share the fac­tor 55: con­tra­dic­tion.
  5. From 00 mark 33 on the line; at 33 raise a per­pen­dic­u­lar of length 11; the hypotenuse from 00 is 9+1=10\sqrt{9 + 1} = \sqrt{10}; swing it down to the line with a com­pass.
  6. 113=0.076923‾\displaystyle \dfrac{1}{13} = 0.\overline{076923}; 213=0.153846‾\displaystyle \dfrac{2}{13} = 0.\overline{153846}. Both blocks have six dig­its; the sec­ond is not a rota­tion of the first, so 113\displaystyle \dfrac{1}{13}'s block is not cyclic in the way 17\displaystyle \dfrac{1}{7}'s is, though 313=0.230769‾\displaystyle \dfrac{3}{13} = 0.\overline{230769} is a rota­tion of the first block.
  7. No: 227=3.142857‾\displaystyle \dfrac{22}{7} = 3.\overline{142857} is ratio­nal and repeats, while π\pi is irra­tional; they agree only to two dec­i­mal places.