I would like some examples of the binary representation of 2-adic integers that are not standard integers. What is the 2-adic expansion of $1/3$? Of $-1/3$? What number does $...010101$ represent?
2 Answers
We can write arbitrary rationals in their $p$-adic expansions by utilizing modular arithmetic and geometric series. To begin with, let's reverse-engineer the form of $p$-adics with periodic digits.
Suppose $0\le a <p^s$. Then
$$n+p^r\sum_{n\ge1}ap^{sn}=n+\frac{ap^r}{1-p^s} $$
Bottom line: we want a denominator of the form $1-p^s$ and a numerator with $p$-free part that is less than the power $p^s$ appearing in the denominator.
We will see it suffices to consider negative rationals $-1<a/b<0$ with $p\nmid b<0$.
To get $1-p^s$ from the $b$ in $a/b$, we employ modular arithmetic. Since $p\nmid b$ we know $p$ is a unit modulo $b$ hence $p^s\equiv1\bmod b$ for some $s$. Let $1-p^s=bc$. Then
$$\frac{a}{b}=\frac{ac}{bc}=\frac{ac}{1-p^s}=ac+acp^s+acp^{2s}+\cdots $$
Furthermore $ac<p^s\Leftarrow ac-1<-bc\Leftarrow (a+b)c<1\Leftarrow a<-b$ so this gives a full expansion.
Now let's re-examine our sufficiency hypothesis. If $x=x_0+x_1p+\cdots$ is a negative rational,
$$-x=(p-x_0)+(p-x_1-1)p+(p-x_2-1)p^2+\cdots $$
is the complete $p$-adic expansion of the positive rational $-x$. Thus we're safe considering negative rationals. Furthermore given arbitrary $x\in\Bbb Q\setminus \Bbb Z$, we can use a combination of subtracting positive integers and negation in order to get $x$ in the interval $(0,1)$, after which we find the $p$-adic expansion and then add the integers back on (which will either perturb a finite number of digits or telescope in an obvious way, e.g. $1+(1+2+2^2+\cdots)$ in $\Bbb Q_2$).
Finally, powers of $p$ can be factored out of a rational and put back in after the expansion of the $p$-prime part has a $p$-adic expansion; powers of $p$ simply shift the digits left or right.
Let's apply this to a concrete example, say $-26/21$ in $\Bbb Q_3$. Factor the $3$ out to obtain $-26/7$, then negate to obtain $26/7$, then subtract $3$ to obtain $5/7$, then negate again for $-5/7$. Calculate powers to obtain $3^6\equiv1$ mod $7$ and $1-3^6=-7\cdot104$. Therefore compute
$$-\frac{5}{7}=\frac{\phantom{-}5\cdot104}{-7\cdot104}=\frac{520}{1-3^6}=520+520\cdot3^6+520\cdot3^{12}+\cdots.$$
As $520_{10}=201021_3$ the above is $\cdots\overline{201021}_3$. Hence
$$\begin{array} -\frac{26}{21} & =-\frac{1}{3}\left(3-\cdots\overline{201021}_3\right) \\ & =-\frac{1}{3}\left(11_3+\overline{021201}021202_3\right) \\ & =-\overline{021201}02122_3 \\ & =\overline{201021}20101_3. \end{array}$$
- 151,657
-
Thanks. I am asking these crazy questions because I am trying to find a FO statement that allows me to distinguish the standard integers from the other 2-adic integers. I shouldn't be able to find one. – Russell Easterly Oct 19 '13 at 01:55
-
1Of course ordinary positive integers have finite $2$-ary expansions. The negative of such a thing goes all $1$’s after a certain point. I guess you knew that, and these are not FO descriptions? – Lubin Oct 19 '13 at 02:44
-
I don't think so. I can't say "if there are an infinite number of 0's". I could say "there are less than n 1's". This only lets me say things about integers of a certain size. – Russell Easterly Oct 19 '13 at 03:37
-
If it helps, every rational integer in $\Bbb Z$, positive or negative, has a unique finite expansion as a polynomial in $\color{Red}{-}p$ with coefficients from ${0,\cdots,p-1}$. – anon Oct 19 '13 at 05:07
-
I can't define "finite" in first order PA. If I could I could prove there are no non-standard models. I would need a FO statement that is true in the rings $\mathbb(Z) /n \mathbb(Z)$ but false in $Z_2$. I was hoping my even or odd statement would work. – Russell Easterly Oct 19 '13 at 16:46
-
@RussellEasterly Why don't you ask that as a separate question? – Keshav Srinivasan Oct 19 '13 at 23:28
To answer your last question first, $$...010101 = 1 + 4 + 16 + \cdots + 4^n + \cdots = 1/(1-4) = -1/3.$$ This also answers your second question. As for your first question, $$1/3 = \dfrac{1}{2}( 1 - 1/3) = ...0101011.$$
- 123,735
-
-
@RussellEasterly: Dear Russell, Yes. Multiplying by $2$ just shifts all the digits to the left (like multiplying by $10$ in usual decimal notation). Cheers, – Matt E Oct 19 '13 at 00:19
-
@MattE would you mind telling me how I can write the p-adic valuation of a number? I read a Wikipedia article on it and since I was a beginner at number theory (saying the truth, I am fiddling around with basics of modular arithmetic), I didn't understand a bit of what the article said on the 5-adic valuation of ${{1}\over{3}}$ and edited a part of it (I dunno if I ever thought it right). – Spectre Jun 26 '21 at 15:58