Video 1, chapter 18

Why 0.999... Equals 1

Why the repeating decimal 0.999... names exactly the real number 1, explained by the limit of a sequence of finite truncations whose gaps from 1 become smaller than any positive number.

  • Video 1, chapter 18
  • When infinity changes the rules

Video 1, chapter 18

Nothing loads from YouTube until you press play.

The numbers 0.90.9, 0.990.99 and 0.9990.999 are all less than 11, by one tenth, one hundredth and one thousandth. So it seems that 0.999…0.999\ldots, with nines forever, should also fall a tiny bit short of 11. Is there any real number between 0.999…0.999\ldots and 11? This page explains what an infinite decimal means, shows that no such number exists, and concludes that 0.999…0.999\ldots and 11 are two names for the same real number.

Key terms

0.999… equals 1

Definition The equality 0.999…=10.999\ldots = 1 says that the repeating decimal with the digit 99 in every place after the decimal point names exactly the real number 11. A decimal with infinitely many digits has no last digit. It is defined as the limit of its finite truncations, which here are

an=0.99…9⏟n nines=1−10−n.\begin{aligned} a_n &= 0.\underbrace{99\ldots9}_{n \text{ nines}} \\[0.4em] &= 1 - 10^{-n}. \end{aligned}

Every truncation is less than 11, and each added nine makes the gap 10−n10^{-n} ten times smaller. The limit is 11 for two reasons. No truncation is above 11, so the limit is not above 11. And no number 1−d1 - d with d>0d > 0 can be the limit, because once 10−m<d10^{-m} < d, the truncation ama_m and all the later ones stay above 1−d1 - d by at least the fixed positive amount d−10−md - 10^{-m}, so they never come close to 1−d1 - d. So 0.999…=10.999\ldots = 1 exactly, in the real numbers. The surprise comes from treating 0.999…0.999\ldots as a very long finite decimal, from picturing a last nine, or from imagining an infinitely small positive gap. None of these exists among the real numbers, and the equality has nothing to do with rounding.

In plain words The decimal 0.999…0.999\ldots is not a list of nines that is still being written. It names the one number that its truncations close in on, and that number is 11.

Example Suppose someone proposes that 0.999…0.999\ldots is one millionth less than 11, so that it equals 1−10−6=0.9999991 - 10^{-6} = 0.999999. The truncation with seven nines is 0.9999999=1−10−70.9999999 = 1 - 10^{-7}, which is already above that number, and so is every later truncation. The same reasoning with more nines defeats any proposed positive gap, however small. An algebra check agrees. Let x=0.999…x = 0.999\ldots. Moving every digit one place to the left multiplies by 1010, and the digits after the point are still all nines, so

10x=9.999…=9+x,9x=9,\begin{aligned} 10x &= 9.999\ldots = 9 + x, \\[0.4em] 9x &= 9, \end{aligned}

and x=1x = 1. This step uses the fact that xx is already a definite real number, the limit above. Subtracting xx from 10x10x removes the whole infinite tail at once and never has to match up a last digit, since there is none.

Repeating decimal

Definition A decimal expansion 0.d1d2d3…0.d_1 d_2 d_3 \ldots, in which each digit dkd_k is one of 0,1,…,90, 1, \ldots, 9, names the real number

d110+d2100+d31000+⋯ ,\frac{d_1}{10} + \frac{d_2}{100} + \frac{d_3}{1000} + \cdots,

which is the limit of its finite truncations. A whole number part in front, as in 9.999…9.999\ldots, is added to this sum. A repeating decimal is one whose digits, from some place on, consist of one fixed block repeated forever. It is written with a bar over the block, as in 0.3‾0.\overline{3} or 0.16‾0.1\overline{6}, or with dots. The repeating decimals, which include the terminating ones that end in repeating zeros, name exactly the rational numbers, the fractions p/qp/q of whole numbers with q≠0q \ne 0. A real number has two decimal expansions exactly when it has a terminating one with a last nonzero digit, and then the other one ends in repeating nines, as in 0.25=0.24999…0.25 = 0.24999\ldots and 1=0.999…1 = 0.999\ldots. Every other real number has only one decimal expansion.

In plain words A repeating decimal is a pattern of digits that goes on forever with no last digit, and it stands for one exact number, the one that its cut-off versions approach.

Example The fraction 1/31/3 has the expansion 0.333…0.333\ldots, because its truncations 0.30.3, 0.330.33 and 0.3330.333 fall short of 1/31/3 by 1/301/30, 1/3001/300 and 1/30001/3000, gaps that shrink toward 00. Multiplying the exact value by 33 gives 3⋅13=13 \cdot \tfrac13 = 1. Multiplying every term of the convergent series 0.333…0.333\ldots by 33 multiplies its sum by 33 and gives 0.999…0.999\ldots, which is another view of the same equality. The decimal 0.24999…0.24999\ldots splits into 0.240.24 and the tail 0.00999…0.00999\ldots, a geometric series whose sum is

91000+9104+⋯=9/10001−110=1100,\begin{aligned} &\frac{9}{1000} + \frac{9}{10^4} + \cdots \\[0.4em] &\quad = \frac{9/1000}{1 - \frac{1}{10}} = \frac{1}{100}, \end{aligned}

so 0.24999…=0.24+0.01=0.250.24999\ldots = 0.24 + 0.01 = 0.25.

Limit of a sequence

Definition A sequence a1,a2,a3,…a_1, a_2, a_3, \ldots of real numbers has the limit LL, written lim⁡n→∞an=L\lim_{n \to \infty} a_n = L, if for every positive number ε\varepsilon there is a whole number NN such that ∣an−L∣<ε\lvert a_n - L \rvert < \varepsilon for every n≥Nn \ge N. A sequence has at most one limit. If every term is at most a number MM and the sequence has a limit, then the limit is also at most MM. If the sequence is increasing and has a limit, then the limit is at least as large as every term.

In plain words Pick any tolerance, as small as you like. From some point on, every term is within that tolerance of the limit and stays within it. The limit is a single number, and it need not be one of the terms.

Example For an=1−10−na_n = 1 - 10^{-n} and L=1L = 1, the distance is ∣an−1∣=10−n\lvert a_n - 1 \rvert = 10^{-n}. For ε=10−6\varepsilon = 10^{-6} take N=7N = 7, since 10−n≤10−7<10−610^{-n} \le 10^{-7} < 10^{-6} for every n≥7n \ge 7, while N=6N = 6 fails, because 10−610^{-6} is not less than itself. For any ε>0\varepsilon > 0 there is a power 10−N10^{-N} below it, and then every later term is within ε\varepsilon of 11. So

lim⁡n→∞(1−10−n)=1,\lim_{n \to \infty} \left(1 - 10^{-n}\right) = 1,

although none of the terms equals 11. The same kind of limit gives the meeting time in Zeno’s Achilles and the Tortoise.

Real number

Definition The real numbers are the numbers that measure positions on a number line. They include the whole numbers, the fractions and irrational numbers such as 2\sqrt{2} and π\pi, with the usual arithmetic and order. Two of their properties matter here. First, every increasing sequence of real numbers whose terms all stay below some fixed number has a limit, and that limit is a real number. This is the completeness of the real numbers, and it is why every infinite decimal names a real number. Second, there is no infinitely small positive real number. For every real d>0d > 0 there is a whole number nn with 10−n<d10^{-n} < d. It follows that a real number d≥0d \ge 0 that is at most 10−n10^{-n} for every nn must be 00, and that two different real numbers p<qp < q always have other real numbers between them, such as (p+q)/2(p + q)/2. Some larger number systems, such as the hyperreal numbers, do contain infinitely small positive numbers, but they are not the real numbers, and an expression like 0.999…0.999\ldots would need its own definition there. On this page, as in school mathematics and calculus, decimals name real numbers.

In plain words On the real number line there is no positive gap smaller than every one of 0.10.1, 0.010.01, 0.0010.001 and so on. Two points with no room at all between them are the same point.

Example Let d=1−0.999…d = 1 - 0.999\ldots. The limit of an increasing sequence is at least as large as each term, so 0.999…≥1−10−n0.999\ldots \ge 1 - 10^{-n}, and therefore d≤10−nd \le 10^{-n} for every nn. Since no truncation is above 11, also 0.999…≤10.999\ldots \le 1, so d≥0d \ge 0. The only real number with both properties is 00, so 0.999…=10.999\ldots = 1. This also answers the opening question. A real number r<1r < 1 is below the truncation 1−10−n1 - 10^{-n} as soon as 10−n<1−r10^{-n} < 1 - r, so it is below 0.999…0.999\ldots too, and no real number lies strictly between 0.999…0.999\ldots and 11.

Finite truncation

Definition The truncation after nn places of a decimal keeps its whole number part and its first nn digits after the decimal point and drops all the later digits. For a decimal 0.d1d2d3…0.d_1 d_2 d_3 \ldots it is the finite decimal

tn=d110+d2100+⋯+dn10n.\begin{aligned} t_n &= \frac{d_1}{10} + \frac{d_2}{100} \\[0.4em] &\quad + \cdots + \frac{d_n}{10^n}. \end{aligned}

Each truncation is a rational number that can be written as a fraction with denominator 10n10^n. For a decimal without a minus sign that names the number xx, the truncations satisfy tn≤x≤tn+10−nt_n \le x \le t_n + 10^{-n}, because the dropped digits add at most

910n+1+910n+2+⋯=10−n,\begin{aligned} &\frac{9}{10^{n+1}} + \frac{9}{10^{n+2}} + \cdots \\[0.4em] &\quad = 10^{-n}, \end{aligned}

and the upper bound is reached only when every dropped digit is 99. For every nn there is a next truncation tn+1t_{n+1}, so there is no last truncation, just as there is no last digit.

In plain words A truncation is what you get by stopping a decimal after some number of digits. It is never above the full value, and it falls short by at most one unit in its last kept place.

Example The truncations of 0.999…0.999\ldots are 0.90.9, 0.990.99, 0.9990.999 and so on, with gaps 0.10.1, 0.010.01 and 0.0010.001 from 11. Every statement about a truncation, such as 0.999<10.999 < 1 or 3⋅0.333=0.9993 \cdot 0.333 = 0.999, is true, and none of them contradicts 0.999…=10.999\ldots = 1, which is a statement about the limit. For π=3.14159…\pi = 3.14159\ldots, the truncations after two and four places are 3.143.14 and 3.14153.1415, and the bound says that π\pi lies between 3.14153.1415 and 3.14163.1416.

Check yourself

  1. Show that 0.4999…0.4999\ldots equals 1/21/2, once with a geometric series and once by finding the gaps between 1/21/2 and the truncations.
Show the answer

The decimal is 0.40.4 plus the tail 0.0999…0.0999\ldots, a geometric series with first term 9/1009/100 and ratio 1/101/10, so

0.4999…=410+9/1001−110=410+110=12.\begin{aligned} 0.4999\ldots &= \frac{4}{10} + \frac{9/100}{1 - \frac{1}{10}} \\[0.4em] &= \frac{4}{10} + \frac{1}{10} = \frac{1}{2}. \end{aligned}

The truncations 0.490.49, 0.4990.499 and 0.49990.4999 fall short of 0.50.5 by 0.010.01, 0.0010.001 and 0.00010.0001. After nn places the gap is 10−n10^{-n}, which becomes smaller than any positive number, and no truncation is above 1/21/2, so the limit is 1/21/2. In the same way every terminating decimal with a last nonzero digit has a second expansion, for example 0.25=0.24999…0.25 = 0.24999\ldots and 3=2.999…3 = 2.999\ldots.

  1. Someone says that the difference 1−0.999…1 - 0.999\ldots is 0.000…10.000\ldots1, a decimal with infinitely many zeros followed by a 11. What is wrong with this?
Show the answer

Each digit of a decimal sits in a numbered place, the first, second, third and so on after the decimal point. A 11 that comes after infinitely many zeros would need a place after every numbered place, and there is no such place, so 0.000…10.000\ldots1 is not a decimal at all. If every place holds a 00, the decimal is 0.000…=00.000\ldots = 0.

The difference can also be found directly. Since 0.999…0.999\ldots is at least each truncation 1−10−n1 - 10^{-n}, the difference 1−0.999…1 - 0.999\ldots is at most 10−n10^{-n} for every nn, and it is not negative. The only such real number is 00, so the difference is exactly 00.

  1. In base two, the digits after the point count halves, quarters, eighths and so on, so 0.111…0.111\ldots in base two means 12+14+18+⋯\tfrac12 + \tfrac14 + \tfrac18 + \cdots. Show that it equals 11, find the gap between 11 and the truncation after nn digits, and find how many digits bring that gap below one thousandth.
Show the answer

The truncation after nn digits is

12+14+⋯+12n=1−12n,\begin{aligned} &\frac{1}{2} + \frac{1}{4} + \cdots + \frac{1}{2^n} \\[0.4em] &\quad = 1 - \frac{1}{2^n}, \end{aligned}

so its gap from 11 is 1/2n1/2^n. These gaps become smaller than any positive number, and no truncation is above 11, so the limit is 11, by the same argument as for 0.999…0.999\ldots. The geometric series formula with a=r=1/2a = r = 1/2 gives the same sum, 1/21−1/2=1\tfrac{1/2}{1 - 1/2} = 1.

Since 29=5122^9 = 512 and 210=10242^{10} = 1024, the gap after 99 digits is 1/5121/512, more than one thousandth, and after 1010 digits it is 1/10241/1024, less than one thousandth. So 1010 digits are needed. In base two the number 11 has the two expansions 1.000…1.000\ldots and 0.111…0.111\ldots, just as it has 1.000…1.000\ldots and 0.999…0.999\ldots in base ten, and in each base the repeated digit is the largest digit available.

How this lesson connects