The repeating decimal \(0.999\ldots\) is not a number that stops just short of \(1\). It is an infinite decimal, and an infinite decimal is defined by a limit. Under that definition,
\[ 0.999\ldots=1. \]
This does not say that every finite string of nines equals \(1\). For every finite \(n\), the decimal with \(n\) nines is strictly smaller than \(1\). The equality appears only after taking the limit of the entire sequence of finite truncations.
What the Ellipsis Means
For each positive integer \(n\), define
\[ s_n=0.\underbrace{99\ldots9}_{n\text{ digits}}. \]
The first few terms are
\[ s_1=0.9,\qquad s_2=0.99,\qquad s_3=0.999,\qquad\ldots \]
The notation \(0.999\ldots\) means the limit of this sequence:
\[ 0.999\ldots:=\lim_{n\to\infty}s_n. \]
Each truncation has the exact form
\[ s_n=\frac{9}{10}+\frac{9}{10^2}+\cdots+\frac{9}{10^n}=1-\frac{1}{10^n}. \]
The gap between \(s_n\) and \(1\) is therefore
\[ 1-s_n=10^{-n}. \]
The gap is positive for every finite \(n\), but it tends to zero. Consequently,
\[ 0.999\ldots =\lim_{n\to\infty}\left(1-10^{-n}\right) =1-0 =1. \]
A Direct Epsilon Proof
The phrase "the gap tends to zero" can be made exact. Given any \(\varepsilon>0\), choose an integer \(n>\log_{10}(1/\varepsilon)\). Then \(10^{-n}<\varepsilon\), and hence
\[ |s_n-1|=10^{-n}<\varepsilon. \]
No matter how small a positive tolerance is requested, every sufficiently long finite truncation lies within that tolerance of \(1\). This is precisely the definition of \(s_n\to1\). There is no positive distance left over after the limit is taken.
Proof by a Geometric Series
The repeating decimal can also be read as an infinite series:
\[ 0.999\ldots =\frac{9}{10}+\frac{9}{10^2}+\frac{9}{10^3}+\cdots =9\sum_{k=1}^{\infty}10^{-k}. \]
This is a geometric series with first term \(9/10\) and ratio \(1/10\). Since the absolute value of the ratio is less than one, its sum is
\[ 0.999\ldots =\frac{9/10}{1-1/10} =\frac{9/10}{9/10} =1. \]
This proof and the limit proof are the same argument in different notation: an infinite series is defined as the limit of its finite partial sums.
Why the Familiar Algebra Works
Let \(x=0.999\ldots\). Multiplying by ten shifts the decimal expansion:
\[ 10x=9.999\ldots \]
and subtracting \(x\) gives
\[ 9x=9, \qquad x=1. \]
The calculation is valid, but it should not be treated as the definition of an infinite decimal. Its legitimacy comes from limits: multiplication by ten and subtraction preserve limits of convergent sequences. Applied first to the finite truncations and then passed to the limit, the apparently informal digit shift becomes an ordinary limit law.
The One-Third Argument
Long division gives
\[ \frac{1}{3}=0.333\ldots \]
so multiplying both sides by three yields
\[ 1=3\cdot\frac{1}{3}=3\cdot0.333\ldots=0.999\ldots \]
This is a useful intuition, but it ultimately relies on the same limit interpretation of repeating decimals. It is not an independent substitute for defining what the ellipsis means.
There Is No Last Nine
A common objection imagines \(0.999\ldots\) as having a final digit after which a tiny gap remains. An infinite decimal has no final digit. If the expansion stopped after \(n\) places, its value would be \(1-10^{-n}\), and the remaining gap would indeed be \(10^{-n}\). But for every proposed final position there is another nine after it.
Nor can the difference be a positive number smaller than every positive real number. If \(d=1-0.999\ldots\) were positive, the Archimedean property of the real numbers would provide some \(n\) with \(10^{-n}<d\). Yet the limiting value lies within \(10^{-n}\) of \(1\) for sufficiently long truncations, contradicting the claim that a fixed gap \(d\) remains.
Decimal Expansions Are Not Always Unique
The equality is one instance of a general fact: every terminating decimal has a second representation ending in repeating nines. For example,
\[ 0.5=0.4999\ldots, \qquad 2.75=2.74999\ldots, \qquad 1.000\ldots=0.999\ldots. \]
Indeed, for any integer \(m\) and positive integer \(n\),
\[ \frac{m}{10^n} =\frac{m-1}{10^n}+\sum_{k=n+1}^{\infty}\frac{9}{10^k}. \]
The two digit strings are different representations of the same real number, just as \(1/2\) and \(2/4\) are different fractional representations of the same rational number. Equality concerns the represented value, not the spelling used to write it.
The Same Phenomenon in Every Base
In positional base \(b\ge2\), the largest digit is \(b-1\), and
\[ 0.\overline{(b-1)}_b =\sum_{k=1}^{\infty}\frac{b-1}{b^k} =\frac{(b-1)/b}{1-1/b} =1. \]
Thus \(0.111\ldots_2=1_2\) in binary and \(0.FFF\ldots_{16}=1_{16}\) in hexadecimal. The duplicate representation is a structural feature of positional notation, not a peculiarity of decimal arithmetic.